ASTERIXDB-PR2: Add schema extraction pipeline for NL2SQL++ (SchemaContextBuilder)

[pineappleBest123]

Summary

Add the schema extraction pipeline for the GSoC 2026 NL2SQL++ project.
This patch builds on top of the servlet infrastructure introduced in
#46.

Changes

• ColumnInfo: field name, type string, and primary-key flag with
  prompt-ready toDescriptionString() output
• DatasetSchema: holds all columns, supports pruned column subset
  (for ColumnPruner in a later PR) and value hints (for ValueHintsSampler)
• DatasetSchemaFormatter: recursively converts ADM IAType objects to
  human-readable strings (supports nested records, arrays, multisets,
  nullable unions, depth limit of 4)
• SchemaContextBuilder: reads Dataset and type metadata from
  MetadataManager, builds a SchemaContext with one description
  string per Dataset, wrapped in a metadata transaction
• 13 unit tests covering all formatter rules and schema pipeline behavior

Example output

Dataset TweetMessages (tweetid: int64 [PK], sender-location: any,
send-time: datetime, referred-topics: [string], message-text: string,
author-id: int64)

Testing

All unit tests pass: mvn test -pl asterixdb/asterix-spidersilk