Linear max_nesting skips deep epics silently: no warning at creation, no doctor check, and the sync never converges

## Summary

Epics deeper than `max_nesting` are skipped by the Linear bridge without any warning outside dry-run output. They never reach the tracker and keep the pending count above zero forever.

## What happened

Six epics sit three levels deep under a `max_nesting: 2` config. Every `tbd --dry-run integration sync` prints:

```
linear: would push 6, would pull 1, skipped 6, warnings 5
- p-aaaa: nested 3 levels, past max_nesting 2
…
```

- Nothing warned when they were created or re-parented to that depth.
- `tbd integration status` and `tbd doctor` report nothing.
- After a full push that updated 173 items, the dry run still says `would push 6`, so the sync can never report converged (compounds #265).

## Suggested change

- Warn at `tbd create --parent` / `tbd update --parent` when the result exceeds the configured depth for an enabled tracker.
- A doctor check listing beads currently past the depth.
- An option to push deep epics as top-level tracker issues with the parent named in the managed description region, instead of dropping them.
- Exclude "skipped" items from `would push` so convergence is readable.

Environment: tbd 0.8.1, Node v24.19.0, macOS 26 (Darwin 25.5), Linear integration enabled. Repository scale when observed: 7,156 beads, 137 open epics, 171 tracker links, about 110 plan specs in `active/`.

Start by reproducing the behavior with `tbd --dry-run integration sync`, then inspect the `tbd create --parent`, `tbd update --parent`, `tbd integration status`, and `tbd doctor` entry points. Done should cover warnings for over-depth parents, a doctor result, an explicit deep-epic handling choice, and convergence output that no longer counts skipped items as pending pushes.

Written by the indexing model from the issue text.