Support embedding batch jobs

[cswaney]

Add support for the TigerFlow-ML embedding task.

[cswaney]

Blocked by #451 (container version bump).

Implementation is complete on feat/embed-batch-task (4c4c2bf) but not opened as a PR, because the task cannot run yet:

• tigerflow-ml's embed task exists only on main (added in feat: add embed task for text embeddings tigerflow-ml#178; image/PDF inputs in Slurm profile using localhost still needs to use profile home directory #180).
• It is not in v0.1.1, the tag pinned in lib/src/blackfish/server/images.py.
• tigerflow-ml publishes images only on release (docker.yml triggers on release: published), so no published image contains embed.

What's on the branch

• jobs/tasks.py — registers embed → tigerflow_ml.text.embed.local, input .txt, output .npy
• asgi.py — COMPATIBLE_PIPELINES["embedding"] → sentence-similarity + feature-extraction
• NewJobModal.jsx — launcher entry, multiPageOnly param gate, applyEmbedEncodeMode
• Unit tests both sides; docs in usage/cli.md and usage/ui.md

907 backend and 521 frontend tests pass; lint and MyPy strict clean.

To unblock

1. Cut a tigerflow-ml release containing embed.
2. Bump the tigerflow_ml tag in images.py (Update container versions #451).
3. Add "embed" to EXPECTED_TASKS in lib/tests/contract/test_tigerflow.py — deliberately omitted, since it would fail against the pinned image. ("chat" is also missing there, a pre-existing gap.)
4. Run the deferred end-to-end check: a job over a directory of .txt files, confirming .npy outputs resolve in the results table.