feat: add loading state, error handling, and tests (closes #2)

[SurbhiAgarwal1]

Summary

Closes #2

Implements loading indicators and structured error handling during the AI context analysis process.

Changes

UI — analyzer.tsx

• Animated multi-step progress bar with named steps:
  Tokenising query → Scanning files → Tracing imports → Scoring relevance → Ranking results
• Skeleton result cards replace the plain spinner during loading
• Classified error messages (network / validation / server / unknown)
  with matching icons and user-friendly text
• Retry button on the error panel — no need to retype the query
• Validation errors show an actionable hint list

App.tsx

• ErrorBoundary catches unexpected render crashes with a recovery screen
• QueryClient configured with retry: false for mutations, retry: 1 for queries

API server — analyze.ts

• Structured error responses with Zod field-level details on 400s
• try/catch around analyzeFiles so runtime errors return a clean 500

Tests

Added 14 Vitest unit tests for analyzeFiles (analyzer.test.ts) — all pass locally.

#	Test
1	returns an array (never undefined/null)
2	returns empty array when no files match
3	ranks most relevant file first
4	filename matches score higher than content-only
5	matchedKeywords populated in results
6	non-empty reason string in every result
7	scores are between 0 and 100
8	respects the topN limit
9	boosts score of files imported by a matched file
10	handles empty files array without throwing
11	handles single-word query
12	case-insensitive keyword matching
13	filters out zero-score files
14	handles files with empty content gracefully

Test run output

[vercel]

@SurbhiAgarwal1 is attempting to deploy a commit to the jarvisx-z's projects Team on Vercel.

A member of the Team first needs to authorize it.

[JarvisX-Z]

Thank you for the excellent contribution. The loading state, structured error handling, and comprehensive unit tests are valuable improvements. I’m reviewing the failed CI build, and if the issue is minor I’ll merge this PR.