feat: add shell completion for analyze command types

[mariusvniekerk]

Summary

• Adds a ValidArgsFunction to the analyze cobra command so that tab-completing the first positional argument suggests the 7 available analyzer types (test-fixtures, duplication, refactor, complexity, api-design, dead-code, architecture) with their descriptions
• After the analyzer type is selected, completion falls back to default file path completion

Test plan

• Run roborev analyze <TAB> and verify all 7 analyzer types appear with descriptions
• Run roborev analyze refactor <TAB> and verify file path completion works
• Verify go build ./... passes

🤖 Generated with Claude Code

[roborev-ci]

roborev: Combined Review (f60e2e8)

Verdict: No Medium, High, or Critical issues found in this PR.

All reviewed findings were either clean or below the reporting threshold for this summary.

---

Synthesized from 4 reviews (agents: codex, gemini | types: default, security)

[mariusvniekerk]

Asked the agent to look for some other opportunities for this.

Flag completions (repeated across many commands):

• --agent — completable from agent.Available() (codex, claude-code, gemini, copilot, opencode, cursor, droid, test)
• --reasoning — just 3 values: fast, standard, thorough
• These flags appear on analyze, fix, run, refine, and review

Positional arg completions:

• fix [job_id...] — could complete with unaddressed job IDs from the daemon (there's already a queryUnaddressedJobIDs() helper)
• repo show/rename/delete/merge — could complete with repo names from the database
• review — could complete git refs/branches

Highest impact, lowest effort would be RegisterFlagCompletionFunc for --agent and --reasoning since they're used on nearly every command and the values are static/easily enumerable.