Pipeline context should properly initialize componentid map when it is created from a proto.

When we re-create context from proto, we don't initialize component_id_map with existing objects in various mappings: https://github.com/apache/beam/blob/b43952ca369418e6f0d6cc2d4c54f991d7aafb25/sdks/python/apache_beam/runners/pipeline_context.py#L184-L211. Therefore we cannot create new entities in the context without stepping over existing IDs.

Imported from Jira [BEAM-12084](https://issues.apache.org/jira/browse/BEAM-12084). Original Jira may contain additional context.
Reported by: tvalentyn.

Start in sdks/python/apache_beam/runners/pipeline_context.py at the proto reconstruction code around lines 184-211. Inspect how existing objects in the mappings are restored and how component_id_map is populated; done means newly created entities do not reuse IDs already represented in the reconstructed context.

Written by the indexing model from the issue text.