Expose table mapping options for MySQL, MongoDB and BigQuery ClickPipes

[sdairs]

Problem and evidence

PR #652 closed #566 for PostgreSQL only and explicitly deferred the analogous MySQL, MongoDB and BigQuery mappings. Their create handlers still parse only --table-mapping source:target and populate required identifiers, defaulting or omitting every advanced mapping field. #593 covers source settings/TLS; #594 covers the streaming destination table; neither covers these database-source mappings.

Scope and command design

Adopt the existing repeatable --table-mapping-json <JSON> on cloud clickpipe create mysql|mongodb|bigquery, combinable with the simple mapping flag using PostgreSQL's ordering and error semantics.

• MySQL: ClickPipeMySQLPipeTableMapping: sourceSchemaName/sourceTable/targetTable plus excludedColumns, useCustomSortingKey, sortingKeys, tableEngine, partitionKey and partitionByExpr added by OpenAPI drift: 108 gaps between live spec and library #522. Explain that partitionKey controls snapshot partitioning while partitionByExpr defines destination PARTITION BY.
• MongoDB: ClickPipeMongoDBPipeTableMapping: sourceDatabaseName/sourceCollection/targetTable and tableEngine.
• BigQuery: ClickPipeBigQueryPipeTableMapping: sourceDatasetName/sourceTable/targetTable plus excludedColumns, useCustomSortingKey, sortingKeys and tableEngine. Preserve omission for the optional fields corrected by OpenAPI drift: 108 gaps between live spec and library #522.

Dependencies and implementation

1. Merge OpenAPI drift: 108 gaps between live spec and library #522 for MySQL partitionByExpr and BigQuery optionality; coordinate with BigQuery ClickPipe create sends zero-valued settings on every request #592's BigQuery settings work.
2. Extend the PostgreSQL JSON-mapping approach with source-specific allowed wire fields and typed enum validation. Reuse mechanics while preserving each source's field names, requiredness and optionality; do not copy PostgreSQL's mandatory default values into the other types.
3. Validate before network/file reads: JSON object, known keys, nonempty required names, known engine, no empty list entries, documented uniqueness, and the sortingKeys/useCustomSortingKey relationship. Keep simple-form request behavior compatible except for optionality fixes intentionally landed in OpenAPI drift: 108 gaps between live spec and library #522.
4. Extract MongoDB and BigQuery pure create builders if necessary. Make mapping helpers reusable by Add clickpipe update for post-create source changes (Postgres/MySQL CDC settings, table mappings, connection) #569 table-add/remove operations without expanding PATCH beyond its schema.

Acceptance and tests

Each source accepts simple-only, JSON-only and mixed input; preserves deterministic ordering; serializes minimal/maximal exact mapping objects; rejects malformed/unknown/missing fields and contradictory sorting flags with usage exit 2 and no request; exercises every supported engine; keeps absent BigQuery options off the wire. Cover MySQL partitionByExpr separately from partitionKey. README gives one meaningful example per source.

Implementation conventions: place clap definitions, request builders, CloudClient wrappers and dispatch in the owning CLI domain; map library errors through CloudClient with organization context. Classify writes exhaustively and fail fast for OAuth writes. Preserve --json/agent output, tolerate missing response fields, and use print_human for detail output. Add README examples. Update both CI classifiers for new files. Validate parsing and minimal/maximal builders, real-binary wiremock auth/request/error/output paths, cargo fmt, default and telemetry-disabled Clippy/check, and tests for every touched crate.