- Replace tea.Program with ports.Renderer in Bridge and OTelTracer
- Update WithProgram() -> WithRenderer()
- Call renderer.OnTaskLog() instead of program.Send()
- Update all tests to use mockRenderer instead of mocking tea.Program
- Fix test assertions to use assert.Positive per linter

- Add OutputMode field to RunOptions
- Remove log file redirection to .same/debug.log
- Implement environment detection and renderer selection (TUI vs Linear)
- Update telemetry setup to use Bridge with ports.Renderer
- Simplify concurrent execution lifecycle (remove manual cancel)
- Remove obsolete TestApp_Run_LogSetupFailure test
- Generate mocks for ports package

- Add --output-mode/-o flag with choices: auto, tui, linear
- Add --ci flag as shorthand for --output-mode=linear
- Wire OutputMode to app.RunOptions
- Default to auto mode (environment-based detection)

Task output now flows exclusively through the OpenTelemetry span →
renderer
pipeline. The logger in the shell executor was creating duplicate
output:
once via slog (unformatted) and once via the renderer (formatted with
[TaskName] prefixes).

Removed:
- Executor.logger field
- logWriter type and methods
- MultiWriter wrapping of stdout/stderr
- Logger dependency from shell adapter node
- Logger mocks from all executor tests

- Add color palette with 8 visually distinct colors
- Implement hash-based color assignment for deterministic task colors
- Apply task color to log prefixes [taskName] in all outputs
- Colorize task name in completion messages while preserving symbol
colors
- Add tests for color assignment consistency and distribution

- Remove logger from shell.NewExecutor()
- Replace tea.Program with Renderer interface
- Simplify MarkExecStart() to only add OTel event
- Update tests to use mockRenderer