perf: speed up percent_decode using bulk run-copying

[jbergstroem]

Speed up ada::unicode::percent_decode, the routine that decodes %XX escapes when parsing hosts (via to_ascii) and search-param values.

The decoded output is never longer than the input, size the result buffer once and copy plain runs in bulk; the same approach form_urlencoded_decode already uses.

What changed

• Allocate the output once at input.size() and shrink to the final length at the end, instead of reserve + per-byte +=.
• Plain (non-%) spans are located with std::memchr('%') and copied with a single std::memcpy, rather than appending one character at a time.
• %XX runs are decoded through the existing unhex_table (two loads + an OR to validate), which is above the function so both percent_decode and form_urlencoded_decode share it.
• Add benchmarks/percent_decode.cpp, mirroring percent_encode.cpp.
• Increase coverage in the test suite to pick up introduced branches.

Benchmark results

Measured with a Apple M5 (benchmarks/percent_decode, --benchmark_min_time=1s):

Benchmark	Before	After	Improvement
Decode time/url	~54.3 ns	~36.4 ns	1.49×
Decode throughput	~1.03 GB/s	~1.54 GB/s	1.49×

Testing

• Full test suite passes
• clang-format, clang-tidy introduces no new findings
• Verified output to be byte-identical to the previous implementation over fuzzed inputs plus a few hand-picked edge cases (%, %2, %%41,%g0, trailing %, doubly-encoded, some utf-8 strings)

Edit: side-note, just noticed #1124 which might be a longer term solution; feel free to close this if thats the preferred path.

[anonrig]

Nice work!

[codecov]

Codecov Report

❌ Patch coverage is 92.59259% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 61.14%. Comparing base (d154358) to head (f6cbe34).

Files with missing lines	Patch %	Lines
src/unicode.cpp	92.59%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1202      +/-   ##
==========================================
+ Coverage   61.06%   61.14%   +0.08%     
==========================================
  Files          38       38              
  Lines        6939     6947       +8     
  Branches     3231     3231              
==========================================
+ Hits         4237     4248      +11     
+ Misses        749      748       -1     
+ Partials     1953     1951       -2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[anonrig]

Edit: side-note, just noticed #1124 which might be a longer term solution; feel free to close this if thats the preferred path.

It has regressions... That's why we couldn't land it.