fix(kanban): keep initially blocked tasks sticky

[hsmon]

Summary

Tasks created with initial_status="blocked" could be immediately promoted by recompute_ready() because creation did not persist sticky blocked lifecycle evidence.

This PR records the initial blocked event atomically with task creation, so operator-created blocked tasks remain blocked until an explicit unblock. Normal dependency-based ready promotion is preserved.

Changes

• Append a blocked lifecycle event during initial blocked task creation in the same transaction.
• Add regression coverage for:
  • initial blocked tasks staying blocked across recomputation;
  • explicit unblock restoring normal promotion behavior;
  • normal ready/dependency paths remaining unchanged.

Validation

uv run --extra dev python -m pytest \
  tests/hermes_cli/test_kanban_blocked_sticky.py \
  tests/hermes_cli/test_kanban_db.py \
  -o addopts= -q

231 passed

Scope

Two files only:

• hermes_cli/kanban_db.py
• tests/hermes_cli/test_kanban_blocked_sticky.py

No gateway restart, config change, migration, credential operation, or destructive cleanup is included.

[alt-glitch]

This was generated by AI during triage.

Duplicate of #34735 (earliest-open canonical for the sticky initial-blocked fix). This PR emits the same explicit blocked lifecycle event in create_task() so _has_sticky_block fires and recompute_ready() won't auto-promote — the identical mechanism, differing only in event payload and test names. Saturated cluster (#46565, #60100 also dup_of #34735).