…on, r=joboet
iter: specialize Take::count using advance_by
Right now, when you call .count() on a Take iterator, it visits every single item one by one just to count them. That's slow.
For example, if you have 1000 items and call .take(5).count(), it should just skip 5 items and return 5 — but today it actually goes through each item manually.
This PR fixes that by using advance_by to skip items in bulk instead of visiting each one. After the skip, we know exactly how many items were consumed, so we return that number directly.
This makes .count() faster for Take iterators, especially when the underlying iterator supports fast skipping.

Rollup of 18 pull requests
Successful merges:
 - #159130 (a bit optimize four-digit chunks in integer formatting)
 - #159592 (core: implement bounded random sampling)
 - #159898 (Add intrinsic-test alias and set  sample rate)
 - #158247 (hermit/fs: Return `unsupported()` instead of `from_raw_os_error(22)`)
 - #158649 (Hermit: fix `readdir()` )
 - #159049 (Avoid ICE in From/TryFrom cast suggestion when encountering HRTBs)
 - #160053 (test: add test suite for the 85681 issue)
 - #160087 (Add regression test for nested associated-type projection ICE)
 - #160090 (rustc_resolve: Further reduce mutability in resolver)
 - #160099 (Resolver: split module resolutions into local and external resolutions)
 - #160106 (Add suggestions for `must_implement_one_of`)
 - #160117 (Remove unnecessary format usage)
 - #160134 (Work around Wine bug 60084 by calling WSAStartup at most once)
 - #160139 (iter: specialize Take::count using advance_by)
 - #160142 (bootstrap: remove use-lld config alias)
 - #160148 (Rename `errors.rs` file to `diagnostics.rs` (15/N))
 - #160151 (Mark a doctest as requiring unwinding)
 - #160166 (Use correct feature gates for `f16`/`f128` `From` impls)

…on, r=joboet
iter: specialize Take::count using advance_by
Right now, when you call .count() on a Take iterator, it visits every single item one by one just to count them. That's slow.
For example, if you have 1000 items and call .take(5).count(), it should just skip 5 items and return 5 — but today it actually goes through each item manually.
This PR fixes that by using advance_by to skip items in bulk instead of visiting each one. After the skip, we know exactly how many items were consumed, so we return that number directly.
This makes .count() faster for Take iterators, especially when the underlying iterator supports fast skipping.

…uwer
Rollup of 18 pull requests
Successful merges:
 - #159898 (Add intrinsic-test alias and set  sample rate)
 - #158247 (hermit/fs: Return `unsupported()` instead of `from_raw_os_error(22)`)
 - #158649 (Hermit: fix `readdir()` )
 - #158693 (Add type-check to offload intrinisc calls)
 - #159049 (Avoid ICE in From/TryFrom cast suggestion when encountering HRTBs)
 - #159411 ([rustdoc] Correctly handle output options with --show-coverage)
 - #160053 (test: add test suite for the 85681 issue)
 - #160087 (Add regression test for nested associated-type projection ICE)
 - #160090 (rustc_resolve: Further reduce mutability in resolver)
 - #160099 (Resolver: split module resolutions into local and external resolutions)
 - #160106 (Add suggestions for `must_implement_one_of`)
 - #160117 (Remove unnecessary format usage)
 - #160134 (Work around Wine bug 60084 by calling WSAStartup at most once)
 - #160139 (iter: specialize Take::count using advance_by)
 - #160142 (bootstrap: remove use-lld config alias)
 - #160148 (Rename `errors.rs` file to `diagnostics.rs` (15/N))
 - #160151 (Mark a doctest as requiring unwinding)
 - #160166 (Use correct feature gates for `f16`/`f128` `From` impls)

Rollup of 20 pull requests
Successful merges:
 - #157669 (cfi: add diag mode support)
 - #158247 (hermit/fs: Return `unsupported()` instead of `from_raw_os_error(22)`)
 - #158649 (Hermit: fix `readdir()` )
 - #158693 (Add type-check to offload intrinisc calls)
 - #159049 (Avoid ICE in From/TryFrom cast suggestion when encountering HRTBs)
 - #159411 ([rustdoc] Correctly handle output options with --show-coverage)
 - #160053 (test: add test suite for the 85681 issue)
 - #160087 (Add regression test for nested associated-type projection ICE)
 - #160090 (rustc_resolve: Further reduce mutability in resolver)
 - #160099 (Resolver: split module resolutions into local and external resolutions)
 - #160101 (Add missing `needs-unwind` annotation to `add-spawn-hook-reentrancy-159923` test)
 - #160106 (Add suggestions for `must_implement_one_of`)
 - #160117 (Remove unnecessary format usage)
 - #160134 (Work around Wine bug 60084 by calling WSAStartup at most once)
 - #160139 (iter: specialize Take::count using advance_by)
 - #160142 (bootstrap: remove use-lld config alias)
 - #160148 (Rename `errors.rs` file to `diagnostics.rs` (15/N))
 - #160151 (Mark a doctest as requiring unwinding)
 - #160166 (Use correct feature gates for `f16`/`f128` `From` impls)
 - #160178 (Remove unused `va_start` intrinsic)

Rollup of 20 pull requests
Successful merges:
 - #157669 (cfi: add diag mode support)
 - #158247 (hermit/fs: Return `unsupported()` instead of `from_raw_os_error(22)`)
 - #158649 (Hermit: fix `readdir()` )
 - #158693 (Add type-check to offload intrinisc calls)
 - #159049 (Avoid ICE in From/TryFrom cast suggestion when encountering HRTBs)
 - #159411 ([rustdoc] Correctly handle output options with --show-coverage)
 - #160053 (test: add test suite for the 85681 issue)
 - #160087 (Add regression test for nested associated-type projection ICE)
 - #160090 (rustc_resolve: Further reduce mutability in resolver)
 - #160099 (Resolver: split module resolutions into local and external resolutions)
 - #160101 (Add missing `needs-unwind` annotation to `add-spawn-hook-reentrancy-159923` test)
 - #160106 (Add suggestions for `must_implement_one_of`)
 - #160117 (Remove unnecessary format usage)
 - #160134 (Work around Wine bug 60084 by calling WSAStartup at most once)
 - #160139 (iter: specialize Take::count using advance_by)
 - #160142 (bootstrap: remove use-lld config alias)
 - #160148 (Rename `errors.rs` file to `diagnostics.rs` (15/N))
 - #160151 (Mark a doctest as requiring unwinding)
 - #160166 (Use correct feature gates for `f16`/`f128` `From` impls)
 - #160178 (Remove unused `va_start` intrinsic)

Rollup merge of #160139 - Rachit2323:take-count-specialization, r=joboet
iter: specialize Take::count using advance_by
Right now, when you call .count() on a Take iterator, it visits every single item one by one just to count them. That's slow.
For example, if you have 1000 items and call .take(5).count(), it should just skip 5 items and return 5 — but today it actually goes through each item manually.
This PR fixes that by using advance_by to skip items in bulk instead of visiting each one. After the skip, we know exactly how many items were consumed, so we return that number directly.
This makes .count() faster for Take iterators, especially when the underlying iterator supports fast skipping.