ci: Update bump-version workflow to regenerate snapshot tests

[malhotra5]

Summary

This PR updates the bump-version.yml workflow to automatically regenerate snapshot tests when bumping the version, and adds verification that each file has exactly 1 line changed.

Problem

The CLI displays its version in the splash screen (OpenHands CLI v1.12.2), which is captured in 25 snapshot SVG files used for visual regression testing. Previously, when bumping the version, these snapshots would not be updated, causing snapshot tests to fail.

Changes

1. pyproject.toml Verification (lines 71-82)

Added a check that verifies pyproject.toml has exactly one line changed (the version line). If more lines are changed, the workflow fails with an error showing the diff.

2. Snapshot Tests Update (lines 103-112)

Added a step that runs pytest tests/snapshots/ --snapshot-update to regenerate all snapshot SVG files with the new version.

3. Snapshot Diff Verification (lines 114-138)

Added verification that each updated snapshot SVG file has exactly 1 line changed:

• Loops through all modified SVG files in tests/snapshots/
• Checks that each file has exactly 2 diff lines (+1, -1)
• Fails with details if any file has unexpected changes

4. Enhanced PR Description (line 205)

The draft PR body now shows:

• The exact diff with verification status: ### pyproject.toml (exactly 1 line changed ✅)
• Count of updated snapshot files with verification: (each with exactly 1 line diff ✅)

Example PR Description

When the workflow runs, it creates a PR with this description:

## Summary

This PR bumps the project version to **1.13.0**.

## Changes

### pyproject.toml (exactly 1 line changed ✅)
```diff
-version = "1.12.2"
+version = "1.13.0"

Other changes

• Regenerated uv.lock file
• Updated 25 snapshot test file(s) (each with exactly 1 line diff ✅)

## Testing

- Validated YAML syntax
- Ran `make lint` - all checks passed

---

## 🚀 Try this PR

```bash
uvx --python 3.12 git+https://github.com/OpenHands/OpenHands-CLI.git@ci/bump-version-snapshot-updates

[github-actions]

Coverage Report •

File	Stmts	Miss	Cover	Missing
TOTAL	6526	910	86%

report-only-changed-files is enabled. No files were changed during this commit :)

[all-hands-bot]

Taste Rating: 🟡 Acceptable with critical fixes needed

Verdict: ❌ Needs rework - Error hiding must be fixed before merge

Key Insight: Solves a real problem pragmatically, but the || true hides failures and could commit broken snapshots.

[all-hands-bot]

Taste Rating: 🟡 Acceptable with critical concerns

Verdict: ❌ Needs rework - Brittle validations will cause false failures in legitimate scenarios

Key Insight: You're validating diff line counts instead of outcomes. Parse and verify what matters (version changed correctly), don't count implementation details (how many lines changed).

[all-hands-bot]

Taste Rating: 🟢 Good taste

Verdict: ✅ Ship it

Key Insight: Clean, pragmatic solution that addresses all previous feedback. Validates outcomes (version value) instead of process (diff line counts), trusts tools (pytest) to do their job, and fails fast if something breaks. No over-engineering, no special cases. This is exactly what the previous reviews pushed for - engineering fundamentals over theoretical perfection.

[malhotra5]

OpenHands bot, if you are reviewing this PR and believe that its worth merging please approve the PR as well

[all-hands-bot]

Taste Rating: 🟢 Good taste

Verdict: ✅ Ship it

Assessment: Clean, pragmatic solution. All previous feedback addressed correctly:

• Validates outcomes (version value) instead of process (diff line counts)
• Trusts pytest to regenerate snapshots correctly
• Fails fast if snapshot update breaks (no || true)
• No over-engineering or brittle special cases

The workflow solves a real problem (version appears in 25 splash screen snapshots) with a straightforward approach: run the snapshot update tool and commit the results. If something breaks, CI tests catch it.

Key Insight: This PR demonstrates good engineering taste - it does the necessary work without trying to validate every detail of how it happened. The previous iterations correctly removed brittle validation logic in favor of trusting tools and testing outcomes.