fix(#560): path_glob page fills from glob-filtered results instead of starving below the rank window

[justrach]

Fixes #560.

Problem

handleSearch fetched offset+max_results+1 globally-ranked results and applied path_glob only at render time. An in-glob match ranking below the window yielded the self-contradicting 0 results … (more results — offset=20). Live repro: codedb_search query='searchContent' path_glob='src/explore.zig' → 0 results despite the function being defined there. Same filter-after-cap family as #547/#553.

Fix

Plain-search branch: with path_glob set, escalate the fetch window ×4 (existing 100000 cap) until the in-glob set fills the page or the index is exhausted, then page over the glob-filtered sequence — offset becomes a cursor into in-glob results and has_more reflects in-glob continuation. No behavior change when path_glob is unset (want_count == old fetch_count, paging over fetched as before). scope=/regex= branches share the bug shape; left for a follow-up per the issue.

Failing Test

test "issue-560: path_glob page must not be starved by higher-ranked out-of-glob files" (src/test_search.zig), committed red: 40 tying out-of-glob decoys ahead of src/gold.zig; pre-fix 67/68.

Validation

zig build test: all green (729 tests), including issue-recall path_glob and issue-422 header-count tests.

Note: trivial EOF conflict with #557 in src/test_search.zig if merged second (both append a test at file end).

🤖 Generated with Claude Code

[github-actions]

Benchmark Regression Report

Thresholds: 10.00% and 50,000 ns absolute delta

NOISE means the percentage threshold was exceeded, but the absolute delta was too small to fail CI.

Tool	Base (ns)	Head (ns)	Delta	Abs Delta (ns)	Status
codedb_bundle	106765	111576	+4.51%	+4811	OK
codedb_changes	11488	11622	+1.17%	+134	OK
codedb_context	1168099	1170356	+0.19%	+2257	OK
codedb_deps	327	401	+22.63%	+74	NOISE
codedb_edit	51515	59466	+15.43%	+7951	NOISE
codedb_find	10238	10311	+0.71%	+73	OK
codedb_hot	28815	28564	-0.87%	-251	OK
codedb_outline	35504	36341	+2.36%	+837	OK
codedb_read	19552	17623	-9.87%	-1929	OK
codedb_search	26353	26393	+0.15%	+40	OK
codedb_snapshot	75424	76475	+1.39%	+1051	OK
codedb_status	9616	10210	+6.18%	+594	OK
codedb_symbol	50006	49630	-0.75%	-376	OK
codedb_tree	50103	53127	+6.04%	+3024	OK
codedb_word	12921	13491	+4.41%	+570	OK

[github-actions]

Benchmark Regression Report

Thresholds: 10.00% and 50,000 ns absolute delta

NOISE means the percentage threshold was exceeded, but the absolute delta was too small to fail CI.

Tool	Base (ns)	Head (ns)	Delta	Abs Delta (ns)	Status
codedb_bundle	114105	106079	-7.03%	-8026	OK
codedb_changes	11762	11183	-4.92%	-579	OK
codedb_context	1180050	1201909	+1.85%	+21859	OK
codedb_deps	332	323	-2.71%	-9	OK
codedb_edit	57404	53032	-7.62%	-4372	OK
codedb_find	12576	10338	-17.80%	-2238	OK
codedb_hot	30120	27731	-7.93%	-2389	OK
codedb_outline	35279	35154	-0.35%	-125	OK
codedb_read	17912	17645	-1.49%	-267	OK
codedb_search	26471	25964	-1.92%	-507	OK
codedb_snapshot	75726	70840	-6.45%	-4886	OK
codedb_status	10520	9795	-6.89%	-725	OK
codedb_symbol	51617	49187	-4.71%	-2430	OK
codedb_tree	23742	23827	+0.36%	+85	OK
codedb_word	13606	12649	-7.03%	-957	OK