Optimize resource reload

[danyadev]

Summary

The old flow first built a content manifest by walking and hashing all relevant files, then walked the same directories again to build GuideNH indexes.

The new flow scans directory packs once and builds the indexes directly. ZIP packs keep their cheap size/mtime-based cache, while directory packs are always rescanned because validating their old cache was more expensive than rebuilding the indexes. Markdown files are also only read far enough to parse the leading frontmatter needed for loadPriority, instead of reading the full page during indexing.

The resource-pack traversal is now isolated in GuideResourcePackScanner, while DataDrivenGuideLoader only handles cache orchestration and applying scan results to GuideNH indexes.

DataDrivenGuideLoader.scanAndBuildAll             ~931 ms → ~190 ms
│
├─ GuideResourcePackScanner.scanDirectoryPack     ~347 ms → ~22 ms
│  └─ collectNamespaceLangPaths                   ~345 ms → ~18 ms
│
├─ GuideResourcePackScanner.scanZipPack                     ~128 ms
│
├─ GuideBuilder.build                                       ~28 ms
├─ apply/index/misc                                         ~12 ms
│
└─ ResourcePackContentManifest.capture            ~430 ms → removed
   └─ contentChecksum                             ~377 ms → removed

Before	After
reload 1277ms	reload 554ms

Checklist

• I have tested this PR in DevEnv
• I have tested this PR in Fullpack
• This PR is in compliance with the GTNH AI Policy
• This PR requires another PR in order to merge