Add channel aliasing, sinkless mode, sequence-of-events, column mapping

[adefabian]

Summary

Adds channel aliasing for the KeyValueStoreSolver, plus four supporting changes:
per-table column-name mapping in solvers, an optional sinkless mode for Report,
a new SequenceOfEvents event type, and selector-aware batching in
Report.determine_report.

Changes

• Channel aliasing: new QueryBuilder.channel_with_alias() and
  MeasurementDB.channel_mapping(); KeyValueStoreSolver resolves aliased
  selectors via a configurable channel_mapping table.
• Selector identity: TimeSeriesSelector now exposes selector_id
  (CRC32) and get_selectors() on every node — used to drive batching and
  alias resolution.
• Per-table column-name mapping: _apply_column_mapping lets each solver
  table (container_tags / container_metrics / channel_tags / channel_metrics
  / channel_mapping / channels) translate physical column names to internal
  ones at read time.
• Sinkless Report: Report.sink is now Sink | None; persistence
  paths are guarded by _has_sink so a report can be built and solved
  without writing to UC.
• SequenceOfEvents: new event type that joins overlapping consecutive
  intervals into one sequence.
• Solver API: filter_channel_tags / filter_channel_metrics now take
  (spark, db, container_df, selectors) instead of (spark, query, …).
  Breaks any external solver subclasses.
• Refactor: Report.determine_report orchestration extracted into
  mda_reporting.core.report_utils (also includes a Best-Fit-Decreasing
  bin-packer for selector-aware batching).

Test Plan

• Unit tests added for channel aliasing, selector_id, sequence-of-events,
  column mapping, batching, and sinkless mode (~1500 lines new).
• Manual testing completed (please confirm)
• Documentation updated (if applicable) — N/A

Checklist

• Code follows project style guidelines
• Self-review completed
• No new linter warnings introduced

[tombonfert]

LGTM. Please check if demo notebooks are still working.

[tombonfert]

LGTM