Word-start matches after /, \, -, . or a tab get no separator bonus, so "domain.cs" outranks "src/main.cs" for "main" #92

What's wrong

Fuzzy.cs:280 only treats _ and a space as separators:

prevSeparator = strChar is '_' or ' ';

So a match that starts a path segment, a kebab-case word or a dotted name gets no matchAfterSeparatorBonus (+10). It scores the same as a match buried in the middle of a word. File paths and command names are what fuzzy finders mostly rank, including TextFilter's Fuzzy filter and Rank, so this is where the ranking goes wrong. Tab is also excluded, even though TextFilter now treats all whitespace as a word separator (TextFilter #112).

The XML doc on matchAfterSeparatorBonus, CLAUDE.md and the README all describe the current _/space behaviour. So this is an enhancement request rather than a bug.

Repro

int S(string s, string p) { Fuzzy.Contains(s, p, out int sc); return sc; }

S("src/main.cs", "main")  //  8
S("domain.cs",   "main")  // 10  <- mid-word match outranks the path-segment match
S("src_main.cs", "main")  // 18
S("src-main.cs", "main")  //  8  (same for "src.main.cs", "src\\main.cs", "src\tmain.cs")
S("get-list",    "list")  // 11
S("getlist",     "list")  // 12  <- outranks the hyphenated word
S("foo-bar", "fb") == S("fxxxbxx", "fb")  // both 5, while "foo_bar" scores 15

Suggested change

Widen the separator test to whitespace plus the common identifier and path separators, for example:

prevSeparator = char.IsWhiteSpace(strChar) || strChar is '_' or '-' or '.' or '/' or '\\';

A small static set would also work. Then update the matchAfterSeparatorBonus XML doc, the README scoring table and CLAUDE.md.

Acceptance criteria

• Score("src/main.cs","main") > Score("domain.cs","main")
• Score("src-main.cs","main") == Score("src_main.cs","main")
• Score("get-list","list") > Score("getlist","list")
• A tab earns the bonus the same way a space does.
• The existing scoring tests still pass, or are updated only where they pinned the old separator set.