[idea] Contextual compact-sidebar labels and clearer status symbols

[bitbonsai]

idea / problem

The compact sidebar currently shows permanent position numbers for workspaces and agents. Workspace numbers are useful while navigating, but outside Navigate mode I find an initial easier to recognize. Agent numbers add visual weight when status is the information I check most often.

The existing status colors are compact, but distinct static shapes make blocked, working, and settled agents easier to distinguish at a glance.

requested change

This prototype explores two related presentation changes.

Contextual compact-sidebar labels

During normal use:

• Workspaces show their first letter beside the status indicator.
• Agent rows show only their status indicator, without a position number.

During Navigate mode, workspace initials switch back to 1, 2, 3, and so on. The existing number-key shortcuts continue working and become visible exactly when they are actionable.

I imagine this as an opt-in setting, preserving the current presentation by default:

[ui]
sidebar_compact_labels = "contextual" # default: "numbers"

This is related to #2141, while keeping the status indicator visible beside the temporary number.

Clearer static symbols

The prototype refines the existing opt-in symbol style:

[ui]
status_indicators = "symbols"

It uses:

• blocked: ▲
• working: ▸
• done/idle: ✓
• unknown: ·

Expanded and compact sidebars use the same symbols. The default dots style remains unchanged. This builds on the distinct-shape work from #824 without introducing continuous rendering.

Screenshots

why you want this

▸ reads as running or active more clearly than a filled yellow circle. The triangle and checkmark distinguish urgent, active, and settled states by shape, with color as supporting information. Checkmarks also feel lighter in the narrow rail.

The contextual labels preserve shortcut discoverability while making normal-mode workspace identity more recognizable.

I have a working local implementation with focused rendering tests. cargo clippy --all-targets --locked -- -D warnings passes.

Would either of these opt-in changes fit Herdr? I am happy to split the implementation if only one direction fits.