test: cover the least-tested run, render, and package modules

[devin-ai-integration]

Summary

Coverage-only change (no src/ edits). pytest --cov on main reported 91.1% overall with a long tail of near-untested modules; this adds 95 tests targeting the seven weakest ones and lifts the total to 93.5%.

module	before	after
integrations/updates/package.py	31%	100%
cli/dispatch/setup_wizard/terminal.py	37%	85%
cli/render/job_logs.py	50%	100%
services/runs/transcript.py	60%	100%
integrations/runs/process.py	68%	100%
cli/render/jobs.py	69%	100%
cli/render/config.py	70%	100%

New files: tests/test_run_transcript_projection.py, tests/test_render_jobs.py, tests/test_render_config.py, tests/test_package_integration.py, tests/test_setup_wizard_terminal.py. Process-control cases were appended to the existing tests/test_run_process.py rather than adding another file.

Notable behavior now pinned by tests:

• project_run_steps — consecutive same-actor text merges into one assistant step, tool_result mutates the matching tool_use step in place (orphan results create their own completed step), agent/error/warning/done/session events map to their step kinds, and context_usage/unhandled kinds are dropped.
• Render functions — human and --json paths for jobs list/show/cancel/logs/attach, including the no log events and detached-attach paths, plus every RunStatus label; render_config_* for values/entry/set/apply.
• PsutilRunProcessController error paths — AccessDenied during termination resumes the frozen tree before raising, survivors after kill raise did not stop, and freeze_process_tree suspends each process once, skips children that exit mid-walk, and resumes what it froze on failure. Spawners are tested against a fake Popen so no real worker/executor is launched.
• InstalledPackageMetadataProvider / read_direct_url — unknown-package fallback, editable direct-URL parsing, and every malformed-payload branch.
• Wizard terminal helpers run against a real pty: wizard_terminal enters/leaves the alternate screen and restores termios, read_setup_key distinguishes a lone ESC from a full arrow sequence. The remaining uncovered lines there are the ImportError (no-termios) branches.

Why

The lowest-coverage modules are exactly the ones that are painful to debug from a user report: transcript projection, job rendering, and process termination. Locking their behavior down makes future refactors in that area safe to review.

Verification

uv run pytest              # 659 passed
uv run ruff check .        # All checks passed!
uv run --with pytest-cov pytest --cov=src/codealmanac   # 91.09% -> 93.47%

Docs and wiki

• Not applicable.

Notes for reviewers

• tests/test_run_process.py gains a FakeProcess stand-in for the psutil tree-walk edge cases that cannot be produced reliably with real processes; the pre-existing real-process tests are untouched.
• test_setup_wizard_terminal.py monkeypatches sys.stdin to a pty follower; the fixture restores termios and closes both fds.

Link to Devin session: https://app.devin.ai/sessions/0496f5a925d14c40a62d1df692a9d26c
Requested by: @rohans0509

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring