splitting annotate.py into quatitate.py and assign.py

[SamuelKrause133]

• xenium_process quantitate: runs MLM/ULM enrichment scoring on a custom
  marker gene list and/or decoupler built-in resources (panglao, hallmark,
  collectri, dorothea, progeny). Supports --filter-obs to restrict scoring
  to a subset of cells (scores written back with NaN for excluded cells).

• xenium_process assign: reads an obsm score matrix produced by quantitate
  and assigns a cell type label to each cluster using a pluggable strategy
  (default: top_positive; threshold and top_n_vote stubbed for future use).
  Optionally runs per-cluster differential expression.

Core changes:

• annotation.py: add filter_cells_by_obs, run_enrichment_scoring,
  load_preset_resource, assign_clusters, STRATEGY_REGISTRY; refactor
  annotate_with_markers to delegate to new primitives
• cli.py: register quantitate + assign; remove annotate
• Delete xenium_process/commands/annotate.py

Config/docs:

• example_config.toml, projects/PDAC_HIV/config.toml: replace [annotate]
  with [quantitate] + [assign] sections
• makefile: run target updated to 6 steps (quantitate + assign replace annotate)

Tests:

• Add tests/unit/test_quantitate_command.py (10 tests)
• Add tests/unit/test_assign_command.py (9 tests)
• Update test_commands.py, test_annotation.py (10 new core function tests)
• Update functional pipeline and config in