fix: prevent byte/char index mismatch panic in jaro_winkler_distance

[xuanwujian]

Fixes #1047

Description

jaro_winkler_distance mixed character indices (from .chars().enumerate()) with byte indices (str::len() and &str slicing) inside get_matched_characters. When a computed slice boundary fell inside a multi-byte UTF-8 character, &str slicing panics (byte index N is not a char boundary). Any non-ASCII input — e.g. jaro_winkler_distance("ab", "céd") — crashed the caller's process.

Beyond the panic, the score formula's length denominators (str1.len() / str2.len()) and the prefix-matching bound were also byte lengths, so a fix that only silenced the panic would still compute a mathematically wrong Jaro score for any multi-byte input (byte length over-counts relative to character count).

Fix: rewrote the function to operate on Vec<char> consistently instead of byte-indexed &str:

• str1/str2 are converted to Vec<char> once at the top; get_matched_characters now takes &[char] and does all windowing/searching by character position.
• The matched-character "removal" still uses the original placeholder approach (replacing the character rather than deleting it), just on a Vec<char> instead of a String — this preserves the exact existing algorithmic behavior/scores for ASCII input (all 6 pre-existing ASCII test cases produce identical results).
• The Jaro score denominators and the prefix-match bound/loop now use character counts, so results are correct — not just panic-free — for multi-byte input.
• Public signature is unchanged: pub fn jaro_winkler_distance(str1: &str, str2: &str) -> f64.
• Left the two secondary/non-blocking observations from the issue thread (matching-window formula vs. the textbook Jaro definition, and first-occurrence-vs-window find semantics for repeated characters) untouched, as noted in the issue as "not the main bug."

Added 3 regression tests: the exact issue repro ("ab"/"céd", no panic), two fully non-ASCII identical strings (CJK 测试/测试 and accented áé/áé), and a partial multi-byte match (测试/测a) with a hand-verified expected score.

Type of change

• Bug fix (non-breaking change which fixes an issue)

Checklist:

• I ran bellow commands using the latest version of rust nightly.
• I ran cargo clippy --all -- -D warnings just before my last commit and fixed any issue that was found. — cargo clippy (nightly) is clean for the file I changed (src/string/jaro_winkler_distance.rs); running it repo-wide with --all -- -D warnings currently fails due to pre-existing warnings in unrelated files on master, unaffected by this change.
• I ran cargo fmt just before my last commit.
• I ran cargo test just before my last commit and all tests passed.
• I added my algorithm to the corresponding mod.rs file within its own folder, and in any parent folder(s). (n/a — existing algorithm, already registered)
• I added my algorithm to DIRECTORY.md with the correct link. (n/a — existing algorithm, already listed)
• I checked CONTRIBUTING.md and my code follows its guidelines.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.89%. Comparing base (c65d014) to head (4cb644c).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1056   +/-   ##
=======================================
  Coverage   95.89%   95.89%           
=======================================
  Files         396      396           
  Lines       30442    30458   +16     
=======================================
+ Hits        29191    29208   +17     
+ Misses       1251     1250    -1     

☔ 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.