Fix revenue dashboard: tenant data leak, rounding, timezones, DB pool

[vidz231]

• Cache: scope Redis key by tenant (revenue:{tenant}:{property}); prop ids are only unique per tenant, so one client was served the other's data.
• Precision: sum as Decimal and round to cents once; stop per-item rounding and float math.
• Timezones: compute month boundaries in each property's local timezone (calculate_monthly_revenue, previously a stub returning 0).
• DB pool: build URL from DATABASE_URL (supabase_db_* settings don't exist), drop QueuePool (invalid for async engines), make get_session sync, add greenlet. The pool never initialised, so the API silently served hardcoded mock revenue.
• Remove mock-data fallback; DB failures now return a logged 503.
• Reject requests with no tenant (403) instead of "default_tenant".
• Return 404 for unknown/foreign properties instead of 0.00 (not cached).

[sentry]

Bug: RevenueSummary retains a stale error state across propertyId changes because the error state is not reset and the component remains mounted.
_{Severity: MEDIUM}

Suggested Fix

Reset error to null before re-fetching data when propertyId changes inside the effect in RevenueSummary, or pass key={propertyId} to RevenueSummary in Dashboard.tsx to force remounting.

Prompt for AI Agent

Review the code at the location below. A potential bug has been identified by an AI
agent. Verify if this is a real issue. If it is, propose a fix; if not, explain why it's
not valid.

Location: frontend/src/components/Dashboard.tsx#L82

Potential issue: In `RevenueSummary`, the `error` state is set when data fetching fails
but is never reset when `propertyId` changes. Because `Dashboard` renders
`RevenueSummary` without a `key` prop tied to `propertyId`, the component remains
mounted across dropdown property changes. Consequently, once an error occurs, switching
to another property continues displaying the stale error message instead of loading the
newly selected property's data.

_{Did we get this right? 👍 / 👎 to inform future reviews.}