fix: support bigquery array of struct type for dbt

[sahmed1991]

Summary

Problem
datacontract import dbt (BigQuery adapter) failed with Unsupported type ARRAY<STRUCT<>> in bigquery json definition because map_type_from_bigquery only understood bare BigQuery type names, not compound types like ARRAY<STRUCT<...>>, STRUCT, STRING(100), or NUMERIC(10, 2).

Changes

datacontract/imports/bigquery_importer.py

• Added ParsedBigQueryType dataclass modelling base_type, optional params (for STRING(n) / NUMERIC(p, s)), optional element_type (for ARRAY / RANGE), and ordered fields (for STRUCT<...>).
  Added parse_bigquery_type(str) -> ParsedBigQueryType — a proper recursive descent parser that handles arbitrary nesting, empty struct bodies (STRUCT<>), whitespace, anonymous struct fields, and parameterised inner types.
• Added build_schema_property_from_bigquery_type(name, type_str, description, required) — builds a full SchemaProperty tree with nested items (arrays/ranges) and properties (structs), while lifting STRING(n) length to logicalTypeOptions.maxLength and NUMERIC(p, s) precision/scale to customProperties (matching ODCS v3.1.0 rules used elsewhere in the codebase).
• Rewrote map_type_from_bigquery to delegate to the parser and return the logical mapping of the outermost type. Added ARRAY, STRUCT, RECORD, RANGE to the base-type mapping. Public signature is unchanged; existing callers continue to work.

datacontract/imports/dbt_importer.py

• create_field now branches on adapter_type == "bigquery" and, for BigQuery columns, uses the new builder so nested items/properties land in the output data contract. Non-BigQuery adapters keep the previous flat behaviour.
• Extracted the column-level metadata attach step (unique, primary_key, custom_props) into _apply_column_metadata so it works for either build path.

Tests

New file tests/test_import_bigquery_type_parser.py (54 tests) covers:

• parse_bigquery_type: scalars, params, arrays, empty structs, ARRAY<STRUCT<>>, nested structs, deep nesting, whitespace, anonymous fields, and error cases (empty input, unbalanced </(, trailing garbage).
• map_type_from_bigquery: simple types, parameterised types, arrays/structs mapping to array/object, and error handling.
• build_schema_property_from_bigquery_type: scalars, STRING(n) maxLength, NUMERIC(p, s) precision/scale, arrays with items, structs with named/anonymous properties, empty struct, ARRAY<STRUCT<>> (the originally reported case), deeply nested types, and required/description propagation.

*tests/test_import_dbt.py: added 4 end-to-end tests that build an in-memory dbt BigQuery manifest and assert the importer produces the correct nested ODCS structure for ARRAY<STRUCT<>>, ARRAY<STRUCT<field ...>>, STRUCT<...> columns, and a regression guard for scalar columns.

Verification

• 69 tests pass across tests/test_import_bigquery.py, tests/test_import_bigquery_type_parser.py, and tests/test_import_dbt.py (was 15 baseline).
• 242 tests pass across the broader -k "dbt or bigquery" selection (3 unrelated skips), confirming no regressions.
• ruff check and ruff format --check both pass on all touched files.

• Tests pass (uv run pytest)
• Code formatted (uv run ruff check --fix && uv run ruff format)
• Docs updated (if relevant)
• CHANGELOG.md entry added