Fix failing tests in fireworks/core/tests/test_launchpad.py + ruff fixes

[janosh]

fix CI errors after merging #555 (which had green CI). looks like that there are some flaky tests in test_launchpad.py that this PR aims to improve. breakdown of changes:

test_launchpad.py fixes

1. Fixed MongoDB Command API Compatibility (Lines 60, 63)

Issue: TypeError: Database.command() takes 2 positional arguments but 3 were given

Root Cause: pymongo 4.x changed the Database.command() API from accepting multiple positional arguments to requiring a single dictionary argument.

Fix: Updated command calls to use dictionary format:

# Before
client.db.command("dropUser", "my-user")
client.db.command("createUser", "my-user", pwd="...", roles=[...])

# After
client.db.command({"dropUser": "my-user"})
client.db.command({"createUser": "my-user", "pwd": "...", "roles": [...]})

2. Fixed WFLockTest Race Condition (Lines 1147-1154)

Issue: test_fix_db_inconsistencies_completed failed intermittently when fw_id=2 timed out waiting for workflow lock.

Root Cause: The 1-second sleep wasn't sufficient to guarantee fw_id=1 acquired the workflow lock before fw_id=2 started.

Fix: Added active polling to wait for fw_id=1 to reach RUNNING state before launching fw_id=2:

timeout = 10
while timeout > 0:
    fw1 = self.lp.get_fw_by_id(1)
    if fw1.state == "RUNNING":
        break
    time.sleep(0.5)
    timeout -= 0.5

3. Fixed WorkflowFireworkStatesTest Cache Inconsistency (Lines 967-970, 1001-1004)

Issue: test_rerun_timed_fws failed with AssertionError: assert 'RESERVED' == 'READY'

Root Cause: When processes are terminated, fireworks can be left in RESERVED state while workflow cache shows READY. The cache is not automatically updated when reservations occur.

Fix: Before each state sync assertion, unreserve any RESERVED fireworks and refresh workflows:

# Unreserve any RESERVED fireworks left by terminated process
reserved_ids = self.lp.get_fw_ids({"state": "RESERVED"})
for fw_id in reserved_ids:
    self.lp.rerun_fw(fw_id)

# Detect lost runs and refresh workflows
detect_lostruns(expiration_secs=0.5, refresh=True)