feat(workflow): per-task subagent routing and a spawn-plan telemetry event (#241)

## Related Issue

No issue. Eighth step of the subagent routing work. Stacked on #240
(merge #234 → #240 first; until then this diff includes their commits).

## Problem

`AgentDynamicWorkflow` could only spawn homogeneous subagents: one
`subagent_type` and one `model` for every item, and nothing reported how
a spawn plan was resolved.

## What changed

- Input gains `tasks[]` (`{ item, subagent_type?, model?, thinking? }`)
and `defaults.subagent_type` beside the homogeneous `items` shape;
`tasks` and `items` cannot be combined (schema refinement). Each task
resolves through `SessionSubagentRoutingService.resolve()`, so one
workflow can mix profiles, models, and thinking efforts and every plan
carries its own provenance. Profile names come from the session catalog;
model aliases use the same vocabulary as `model`. `model` stays the last
top-level parameter so the pool-less schema strip is unchanged.
- `SubagentSpawnPlanInput.thinking` (explicit effort) is honored by the
resolver and is part of `routeDecisionFingerprint` only — never of
`routingEnvironmentRevision`.
- Telemetry: `subagent_spawn_plan_resolved` (`operation`,
`profile_source`, `model_source`, `policy_mode`, `policy_source`,
`feature_source`, `routing_env_revision`, `route_decision`,
`explicit_profile|model|thinking`). No prompt content.
- Token-count expectations in the compaction/loop tests and the wire
snapshots follow the larger tool schema (verified: the only delta is the
new `defaults`/`tasks` parameters and the tool-list hash).

Tests: mixed-profile workflow (three tasks → three plans with different
profile/model/thinking/provenance); `tasks` + `items` rejected; legacy
`items` shape unchanged; explicit thinking changes the decision
fingerprint but not the environment revision; telemetry attribute set
asserted exactly.

## Checklist

- [x] I have read the
[CONTRIBUTING](https://github.com/PyModel/pythinker-code/blob/main/CONTRIBUTING.md)
document.
- [ ] I have linked a related issue (external PRs: the issue must have a
maintainer's `/approve`).
- [x] I have added tests that prove my feature works.
- [x] Ran `gen-changesets` skill, or this PR needs no changeset.
- [x] Ran `gen-docs` skill, or this PR needs no doc update.


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **New Features**
* Dynamic workflows now support per-task subagent types, models, and
thinking settings, with optional defaults.
* Subagent routing and model selections are preserved when workflows
resume or create child tasks.
  * Added telemetry visibility for resolved subagent spawn plans.
  * Expanded diagram and editor support in the web experience.

* **Bug Fixes**
  * Improved workflow task counts when tasks are still resolving.
* Automatically clears stale secondary-model settings during provider
refreshes.
  * Improved spacing consistency across workflow cards.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Author: elkaix

.changeset/dynamic-workflow-per-task.md

@@ -0,0 +1,5 @@
+---
+"@pymodel/pythinker-code": minor
+---
+
+AgentDynamicWorkflow accepts a `tasks` list where each entry sets its own subagent type, model, and thinking effort. Pass `tasks` instead of `items`, with optional `defaults.subagent_type`.

apps/pythinker-code/dist-web/.web-bundle-manifest.json

@@ -1,4 +1,4 @@
 {
-  "sourceHash": "c294d9482527fbd47119352fcaac10f9870f9ef62d7228be22642ea8bd1a7804",
+  "sourceHash": "1e2f179018f506a81f8da2b82e60a8331dedf5e9683f13ea4a64e2625c0116fc",
   "sourceFileCount": 407
 }