fix: correct storage.update() call signature in _update_active_counts()

[ponsde]

Problem

Session._update_active_counts() has three bugs that leave active_count permanently stuck at 0.

Bug 1: wrong keyword argument names

# Broken call in session.py:
storage.update(
    collection="context",
    filter={"uri": usage.uri},            # ← not a valid kwarg
    update={"$inc": {"active_count": 1}}, # ← not a valid kwarg
)

# Actual signature of VikingVectorIndexBackend.update:
async def update(self, collection: str, id: str, data: Dict[str, Any]) -> bool:

Python raises TypeError: update() got an unexpected keyword argument 'filter' on every call. The exception is silently swallowed by except Exception, so active_count is never written.

Bug 2: unsupported $inc operator

storage.update() is merge-and-upsert ({**existing, **data}). It does not support MongoDB-style increment operators. The value must be a plain integer.

Bug 3: fetch_by_uri returns subtree matches on path fields

Even after fixing bugs 1 and 2, using fetch_by_uri() to locate the record is unreliable. The local VikingDB engine applies hierarchical (subtree) matching to path-type fields, so filtering uri = 'viking://resources/doc' also matches child records like 'viking://resources/doc/chunk1'. When more than one record matches, fetch_by_uri raises ValueError("Duplicate records found") and returns None — active_count still goes unupdated.

Verified empirically:

filter(must, uri="viking://resources/doc")
→ returns ["viking://resources/doc", "viking://resources/doc/chunk1"]   # subtree match

Fix

Compute record_id = md5(uri) directly — the same formula used in collection_schemas.py for every context record — then use storage.get(ids=[record_id]) for a precise single-record lookup unaffected by path-field matching semantics:

import hashlib

record_id = hashlib.md5(usage.uri.encode("utf-8")).hexdigest()
records = run_async(storage.get(collection="context", ids=[record_id]))
if not records:
    continue
current_count = records[0].get("active_count") or 0
run_async(storage.update(
    collection="context",
    id=record_id,
    data={"active_count": current_count + 1},  # plain int, not $inc
))

Verification

All three bugs independently verified (3/3 runs each):

• Bug 1 (mock test): TypeError: update() got an unexpected keyword argument 'filter'
• Bug 2 (source read): update() implementation confirmed as merge, no $inc support
• Bug 3 (empirical test): path filter on must op returns subtrees, not exact matches

End-to-end test result before fix (3/3 runs):

commit.active_count_updated = 0   # never incremented

Changes

• openviking/session/session.py: add import hashlib; replace broken kwargs + $inc + fetch_by_uri with storage.get(ids=[md5(uri)]) lookup and correct update(id, data) call
• tests/session/test_session_commit.py: add regression test using storage.get() for exact id lookup; docstring covers all three bugs

[CLAassistant]

All committers have signed the CLA.