fix(preflight): mirror SequenceGroupValidator's null set for sequence ids (#239)

[LukasWodka]

⚠️ Stacked on #218 — merge that first

This branch is based on the local feat/218-tsc-preflight-parity work (pandasDefaultNA), which is not on develop yet. Until #218 lands, this PR's diff shows two commits:

• fe46ff0 — TSC preflight parity: mid-sequence label-flip + per-group unsorted timestamps pass local dry-run but are rejected in-cluster #218 (the two grouped TSC validators + the pandasDefaultNA map)
• 8105679 — this change (the only commit to review here)

Once #218 is merged to develop, GitHub will drop fe46ff0 from the diff and this PR will show only the sequence-id fix. If #218 ships separately first, rebase this on the updated develop.

---

What

Closes #239. Same family as #218, a different validator not covered by it.

CheckSequenceRows / sequenceScanFrom (internal/push/preflight.go) decided whether a sequence_id cell is null by trimming it and testing the curated coercion.NA_SENTINELS map (naSentinels). But it previews the ingestor's SequenceGroupValidator, which plain-reads the CSV (pd.read_csv, keep_default_na=True) and computes null as:

ids=df[resolved]
null_mask=ids.isna() | (ids.astype(str).str.strip() =="")

So the faithful null set is pandas' default STR_NA_VALUES (matched on the raw cell — pandas tokenises NA before stripping) ∪ whitespace-only cells, not the coercion set. Two divergences:

sequence id	pandas / ingestor	old CLI (naSentinels+trim)	direction
none (lowercase)	real id → accept	in naSentinels → null → reject	over-reject (dangerous) — blocks an ingest the cluster accepts
#NA	in STR_NA_VALUES → null → reject	not in naSentinels → real id → miss	under-reject
" NA " (padded)	not an NA token → real id	trimmed to NA → null	over-reject

Fix

Reuse the pandasDefaultNA map #218 added for the grouped checks:

a sequence_id cell is null iff raw ∈ pandasDefaultNAORstrings.TrimSpace(raw) == "" — matched on the raw cell, dropping the pre-strip.

"" is already in pandasDefaultNA, so the whitespace clause only adds genuine whitespace-only cells. Distinct ids are counted on the raw value (pandas groups the object key verbatim, consistent with #218's labelConstantViolation / perGroupTimeViolation). naSentinels stays in use for the label-diversity read path (which the ingestor genuinely pins to coercion.NA_SENTINELS, keep_default_na=False).

Test plan

• New parity casetsc-none-sequence-id (accept), driven against the realSequenceGroupValidator — goldens.json regenerated at the pinned data-ingestors ref (7b4ecac); the goldens diff is exactly the one new accept entry, and scripts/sync-validator-goldens.sh --check is green.
• Existing tsc-null-sequence-id (empty + NA) still rejects.
• TestCheckSequenceRows gains none / #NA / whitespace-only / padded-NA boundaries, each ground-truthed against the validator's null_mask.
• Mutation-proved: reverting sequenceScanFrom to naSentinels+trim fails the new parity case (over-reject) and the none/#NA/padded assertions (both directions).
• go build / go vet / gofmt clean; go test ./... green; scripts/coverage-floor.sh green (internal/cli 73.9%, internal/submit 76.1%).

🤖 Generated with Claude Code

---

Note

Low Risk
Scoped to local preflight parity for time-series sequence id validation; behavior change only affects edge-case NA/id strings, with tests and parity pins guarding regressions.

Overview
Aligns CheckSequenceRows with ingestor SequenceGroupValidator: null sequence_id cells are detected with pandasDefaultNA on the raw CSV value plus whitespace-only, instead of trimmed coercion.NA_SENTINELS. Distinct sequence counts also use raw ids so grouping matches pandas.

This fixes over-reject when the id is the literal none (pandas keeps it) and under-reject for tokens like #NA, and treats padded values such as " NA " as real ids. Label diversity still uses naSentinels.

Adds unit boundaries and parity case tsc-none-sequence-id with updated goldens.

^{Reviewed by Cursor Bugbot for commit 3bd3fe2. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

Retargeted develop → feat/218-tsc-preflight-parity to reconcile a cross-session snapshot mismatch. This PR was opened on the understanding that cli#218 (which added the pandasDefaultNA map this fix reuses) was unpushed with no PR — but #218 is in fact PR #234 (open, base develop). So this now stacks cleanly on #234 and its diff shows only the #239 sequence-id fix (8105679), not #218's commit.

Merge order: #234 first, then this. When #234 merges to develop (and its branch is auto-deleted), GitHub will retarget this PR to develop automatically. The fix itself is unchanged and green.