implement entity id attr pick-up mech and fix #31

[subbyte]

The PR fully addresses #31 and partially addresses #32.

1. implement a new function get_entity_id_attribute() in src/kestrel/codegen/relations.py to compute the appropriate attribute used as identifier attribute for entities.
2. update src/kestrel/codegen/commands.py to use get_entity_id_attribute().
3. replace the or_pattern() for post-prefetch merge in src/kestrel/codegen/commands.py with firepit.merge() (partially address codegen upgrade: use firepit to merge variables #32).
4. update get_variable_entity_count() in src/kestrel/codegen/summary.py to use get_entity_id_attribute().
5. update _get_variable_query_ids() in src/kestrel/codegen/summary.py since merged variable in firepit do not have __membership records.
6. update gen_variable_summary() in src/kestrel/codegen/summary.py to only give cached records when there is a data source query.

[CLAassistant]

All committers have signed the CLA.

[pcoccoli]

Looks OK, though ideally we should be able to use multiple attributes for object identity (e.g. for a process, use both pid and name).

[subbyte]

Looks OK, though ideally we should be able to use multiple attributes for object identity (e.g. for a process, use both pid and name).

Good point to make the function get_entity_id_attribute() more comprehensive regarding logic in the future.