Fix ADF→Lakeflow classification + single-bundle packaging for multi-pipeline factories

[matthewmoorcroft]

⚠️ Draft — initial port. This change is rebuilt onto the unified base as-is. It predates the engine → sources/adf module restructure, so it still needs adaptation to the new layout (and a green test run) before it is mergeable.

Fixes three flowx bugs surfaced while migrating 6 real Telefónica ADF/Synapse exports end-to-end through discover → convert → package → validate. With these changes, all 6 exports classify at 100% coverage
(0 unsupported) and package into a single deployable bundle each that passes databricks bundle validate with 0 errors.

This branch already carries the fixes from PR the multi-pipeline report fix and PR the bridge-notebook header fix.

Changes:

• ADF exports the activity type as AzureFunctionActivity, but AGENTIC_TYPES only listed AzureFunction, so every Azure Function activity fell through to unsupported (no translation path) instead of agentic. Add "AzureFunctionActivity" to AGENTIC_TYPES.
• Packaging emitted one databricks.yml bundle per pipeline. New _combine_into_single_bundle() folds every pipeline into one workflow tree (reusing the existing inner_workflows machinery write_bundle already uses to emit a parent plus its children as separate job resources). Each pipeline keeps its own job name, parameters, and schedule.
• DAB job parameters require a default; flowx omitted the field for ADF params that had none, so bundle deploy failed with "The argument 'default' is required.". Emit default: "" for params with no source default.

[matthewmoorcroft]

Closing — this was a port of another author's work and is being withdrawn from the public repo.