Roles, capability profiles and help on the new loop (M2.6)

- roles.driver (unset: the planner's model) and roles.classifier (unset:
  the expert's); the mapping is pinned when an item opens, and the driver
  and the go-ahead reader run on the item's own (L9).
- profiles: in configuration (tools, strengths, weaknesses), used in
  briefings: the driver's instructions say what its helper and expert are
  good at, the helper's brief who reads its handback; never a model name.
- delegate(task) and ask_expert(question) for the driver. A helper works
  on a branch off the milestone's; on a handback its work comes back as
  uncommitted changes, otherwise it is discarded, so harness branches
  never hold commits only the clone has. ask_expert keeps its budget.
- Triggers: failure_limit failed commands in a row (Driver's new "stuck"
  outcome) or the turn cap, mid-milestone, call a helper in once per
  milestone; if it fails too, the item waits on the operator. A
  delegation cut off by a crash is discarded at boot.
- Evaluation capture: delegations and ask_expert as before, and the
  reader's verdicts as classifier pairs.

Tests: help tests (merge-back, discard, needs a milestone, the stuck
trigger and its once-per-milestone limit, the expert and its budget,
pinning, briefings by profile, boot discarding a cut-off delegation),
config tests for roles and profiles, and the walk delegating with a
scripted helper that sometimes fails, with a delegations_settle property.
The walk found helper commits stranded on a merged PR's branch, fixed by
returning a helper's work uncommitted.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PPQ14L3LikyefCMFBqEbKE

nisse-dev.example.yaml

@@ -45,6 +45,27 @@ roles:                                # L9: role -> LiteLLM model alias
   planner: claude-opus-5-5      # high-end judgement: plans a weaker executor follows
   helper: claude-opus-5-5     # takes over what the executor got stuck on
   expert: claude-haiku-4-5  # one question at a time, no tools: cheap is enough
+  # The new loop's roles (loop: new): unset, driver is the planner's model and
+  # classifier the expert's. Each work item pins the mapping when it opens.
+  # driver: claude-opus-5-5     # drives the loop: talks, restates, works
+  # classifier: claude-haiku-4-5  # reads plain-language go-aheads (later: actions, L4)
+
+# Capability profiles (L9), from benchmarks: what one role is told about
+# another (never a model's name). Optional; a model without one is briefed as
+# "no profile recorded".
+profiles:
+  qwen3.8-27b:
+    tools: reliable (37/40 on the executor benchmark, PR #55)
+    strengths: careful multi-step edits that follow a clear brief
+    weaknesses: speed (minutes per hard task) and open-ended design
+  claude-haiku-4-5:
+    tools: reliable
+    strengths: quick, well-scoped changes and short answers
+    weaknesses: long multi-file work (15/20 on the harder benchmark tasks)
+  claude-opus-5-5:
+    tools: reliable
+    strengths: design, debugging, and work other models got stuck on
+    weaknesses: cost; use it for what needs it
 
 matrix:                               # L5
   homeserver: https://matrix.example