feat(tui): size side panels by content instead of equal shares

[bug-ops]

Summary

Replaces the TUI dashboard's side-panel column (Skills, Memory, Resources, SubAgents) equal-Fill(1) split with content-driven sizing, so a sparse panel no longer wastes rows and a busy one is no longer silently clipped.

• New layout::{PanelDemand, PanelSizing, fit_panel_heights} — an integer max-min-fair water-filling allocator. PanelDemand::default() == Greedy reproduces the pre-existing equal-share layout's totals exactly (per-slot remainder placement differs from ratatui's cassowary solver — documented, not treated as a bug).
• Each measured widget (skills, memory, resources, subagents' plain list, security, plan_view, compaction_badge) exposes a pure desired_height derived from the same line/row builder its render uses, so measurement and rendering can never drift apart.
• New shared widgets::panel::render_lines truncates to the granted height and shows a muted +N more indicator instead of silently dropping content under pressure.
• App::panel_demands() composes widget measurement with chrome (focused-panel header, resources' gauge/compaction badge, the subagents equalizer carve-out); App::subagent_slot_mode() is now the single source of truth for which view the SubAgents slot shows (previously decided independently in two places).
• New [tui] panel_sizing config (auto/even) wired through the --init wizard, --migrate-config (step 107), the /panel_sizing command, and the command palette.
• The collapse toggle is kept as an explicit user pin (PanelDemand::Collapsed) that overrides content-driven sizing.

Went through a full validation round (tester, perf, security, adversarial critic) before code review; the critic's first pass caught 2 critical + 3 significant issues (an equalizer-never-renders regression on an empty sub-agent list, an off-by-one in the overflow indicator, the skills panel silently dropping its MCP section under pressure, and an over-claimed behavioral-parity docstring) — all fixed and re-verified before this PR was opened.

Test plan

• cargo +nightly fmt --check
• cargo clippy --profile ci --workspace --all-targets --features "desktop,ide,server,chat,pdf,scheduler,testing" -- -D warnings
• cargo nextest run --config-file .github/nextest.toml --workspace --features "desktop,ide,server,chat,pdf,scheduler" --lib --bins — zeph-tui 1067/1067, zeph-config 990/990
• Rustdoc gate (RUSTFLAGS=-D warnings + --deny rustdoc::broken_intra_doc_links)
• cargo test --doc --workspace --features "desktop,ide,server,chat,pdf,scheduler"
• cargo insta test — zeph-tui/zeph-config clean, 0 snapshot diffs
• gitleaks protect --staged
• specs/011-tui/spec.md updated with a Side Panel Sizing section
• CHANGELOG.md [Unreleased] entry
• docs/src (book/src/advanced/tui.md) updated
• .local/testing/playbooks/tui.md scenarios + coverage-status.md row added (main repo)
• Live tmux-pty testing (playbook scenarios) — not yet run, flagged for the tester/live-tester agent as next step

Closes #6675