Add timezone_aware option for continuous aggregates

[jonatas]

Hello Timescale friends, I had a need for time zones and I think it fits perfectly on the continuous aggregates helper.

Please, check if this would be useful for the core library.

Summary

• Adds timezone_aware: true option to continuous_aggregates — injects an in_timezone(tz) scope into every generated cagg class
• Adds timeframe_intervals: to support custom SQL intervals for symbolic timeframe names (e.g. { halfhour: '30 minutes' }) covering X:30 UTC offsets like IST, ACST, IRT, and NST
• The scope rebuckets UTC cagg rows into local calendar days using (time_column AT TIME ZONE tz)::date without extra caggs or raw-table scans
• Wraps the rebucketing in a derived table so chained .where calls work correctly
• Benchmarking showed this scope approach outperforms permanent SQL VIEWs because the time_column predicate stays in the inner query where the planner can apply chunk exclusion

Why sub-daily granularity matters

The daily cagg buckets by UTC midnight, which always straddles local calendar days for non-UTC zones. Reading from a 30-min cagg and rebucketing by local date gives correct results for every timezone — including X:30 offsets.

# EST user sees the 03:00 UTC row correctly on Jan 15, not Jan 16
PageView::PageViewStatsPerHalfhour
  .in_timezone('America/New_York')
  .where(user_id: 2, local_date: '2024-01-15')
  .sum('total_duration')
# => 2700  (03:00 UTC = 22:00 EST on the 15th ✓)

Test plan

• 23 new specs covering class generation, SQL correctness, and data correctness with a day-boundary proof
• Full test suite passes (pre-existing failures unrelated to this change)
• Runnable example script in examples/all_in_one/timezone_aware_caggs.rb
• Docs updated in docs/models.md