TW-4829: Harden CLI calendar scheduling flows and test coverage

[qasim-nylas]

Summary

This PR hardens the CLI calendar and timezone scheduling paths by replacing misleading or stubbed behavior with real scheduling logic, reducing duplicate helper code in touched paths, and restoring test discovery that was previously missing important coverage.

Jira

• TW-4829
• Parent epic: TW-4638

What changed

• wired calendar ai schedule through real event-creation logic instead of reporting success without creating anything
• moved AI scheduling tool behavior onto a real implementation path in internal/adapters/ai/scheduler_tools.go
• replaced placeholder timezone meeting-finder behavior with real overlap computation in the shared timezone service
• removed hard-coded participant timezone assumptions from calendar find-time and aligned it with shared scheduling logic
• fixed minute-precision working-hours scoring so ranges like 09:30-17:30 are handled correctly
• reduced duplicate JSON request/response handling in the touched Nylas adapter methods by routing through shared helpers
• introduced a shared CLI command test helper in internal/cli/testutil/command.go
• renamed previously skipped *_test_basic.go / *_test_advanced.go files to real *_test.go names so go test discovers them normally
• added regression coverage for AI scheduling and minute-based meeting-time scoring
• removed the false-positive security-scan trigger in internal/cli/dashboard/keys.go

Why this matters

Before this change, parts of the scheduling experience could report success without actually performing the underlying action, or return false negatives because the implementation was effectively stubbed. On top of that, some important tests were not being executed at all because of filename conventions, which made regressions easy to miss.

Validation

• make ci-full
• full unit tests
• race detector suite
• security scan
• vuln check
• integration tests
• Air integration tests

Notes

• make ci-full passed on this branch immediately before opening the PR.
• The PR includes the test-file renames that restore discovery, so the diff is larger but most of those changes are mechanical renames rather than logic churn.

[AaronDDM]

LGTM 👍