The first cycle on the cloud PC failed with:
  "Indicators failed for group al_1d: start_date and end_date are
   required when trading_dates is None."

Root cause: phase0_pipeline._calculate_indicators_per_group reads
sc.start_date via getattr with default None. SlotDeployConfig has
no start_date field — getattr always returns None — so start_date
is None, run_indicator_calculation rejects the call, no
strategy_indicators.csv is produced, slot init then fails reading
the missing CSV. 100% repro on first daily cycle on a fresh deploy.

Fix: when no slot in the group provides a start_date, fall back to
present_date - 4 years. 4y warmup is enough for any realistic
indicator parameter; data download cost is small. Slots that need
more (or less) can override by setting start_date as a dataclass
attribute on their slot config — getattr's existing read picks it up.

Logged at INFO level so the operator sees the default kicked in:
  "Indicators: al_1d no slot.start_date provided; defaulting to
   2022-05-08 (present - 1460 days)"

Tests: 173 passed, 1 skipped under tests/live (full live suite).

Bumped to 0.1.6.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>