exp: Propagate column types between nodes

[aMayzner]

Summary

Implements type propagation across query builder nodes to preserve column type information through transformations. Previously, column types were often lost when flowing through nodes (add columns, aggregations, etc.), leading to loss of semantic information like TIMESTAMP, DURATION, etc. The implementation includes proper type inference for aggregations and warnings when type information cannot be determined.

Changes

• AddColumnsNode: Preserves types from source columns when selecting columns and when creating computed columns with stored type information. Logs warnings whenstored type parsing fails.
• AggregationNode: Adds getAggregationResultType() helper to determine correct result types based on operation semantics (COUNT→INT, MEAN/MEDIAN→DOUBLE, SUM/MIN/MAX preserve input type). Logs warnings when type information is missing for operations that should preserve it.
• TimeRangeSourceNode: Explicitly specifies column types (id→INT, ts→TIMESTAMP, dur→DURATION) as a proper source node pattern.
• Type parser: Made table and column parameters optional in parsePerfettoSqlTypeFromString() to support contexts where table/column info isn't available.
  Properly errors when plain 'id' type is parsed without context.
• Tests: Added comprehensive unit tests for type propagation in aggregation and timerange source nodes, covering all aggregation operations and edge cases.
• Debugging: Added console warnings when falling back to default INT type, helping developers identify and debug type propagation issues.

Type Inference Rules

The implementation follows these type inference rules for aggregations:

• COUNT, COUNT_ALL → INT
• SUM, MIN, MAX → Preserves input column type
• MEAN, MEDIAN, DURATION_WEIGHTED_MEAN, PERCENTILE → DOUBLE
• Unknown/missing type → INT (with warning)

[github-actions]

🎨 Perfetto UI Build

✅ UI build is ready: https://storage.googleapis.com/perfetto-ci-artifacts/gh-19786515871-ui/ui/index.html