CellCellConnectivityLong: add connectome_id

Problem: no measurement-context discriminator — rows from two measurement contexts are indistinguishable and their overwrite predicates clobber each other. Hence today's one-subdirectory-per-connectome workaround (cellcellconnectivitylong_*).

Semantics: connectome_id identifies the measurement context (segmentation version + proofreading state + measurement semantics), not the cohort. Cohorts are read-side filters: write the superset connectome once; pull subsets by joining dataitem_dataset_association or cluster membership. Minnie's two folders (proofread_pre_to_csm_post, proofread_to_proofread) are two measurement contexts → two connectome_ids in one table.

Fix: add connectome_id; register a WriteSpec (overwrite_scoped on (project_id, connectome_id)); rerun affected ETLs with the new io + schema (no folder migration); add a subset-read helper (ties to #23). Needs its own planning session before implementation.

Start by locating the CellCellConnectivityLong schema, its io and WriteSpec registration, and the affected ETLs. Review the planned relationship to issue #23 and use the planning session to settle the subset-read helper. Done means the new connectome_id supports scoped overwrites, affected ETLs use the updated schema and io, and no folder migration is required.