Confluence import: store layer and CAS transitions

[Willyfrog]

Layer 4 of 7 in the Confluence page import stack, split out of #18. The design narrative and the accepted risks live in #18 — read it first; this PR is one layer of it.

Stack order: #30 → #31 → #32 → #33 → #34 → #35 → #36

The persistence layer, and where the concurrency design actually lives. Every state transition is a compare-and-set against the state the selection read, so a stale read loses its CAS rather than duplicating work — there are no leases and no heartbeats by design.

What's here

import_store.go, import_page_store.go, import_preflight_store.go, import_terminal_store.go, import_report_store.go.

What to look for

• Is every transition genuinely a CAS? A read-then-write that isn't conditioned on the state that was read is the bug class this design exists to prevent. Look for any UPDATE whose WHERE omits the expected prior state.
• Serialization by selection, not by ordering — a job that may have written pages fences its source and Space until it terminalizes. Check the fence is taken before the first write, not after.
• Progress is reconstructed from immutable per-page checkpoints, never remembered. Confirm nothing caches progress in a way a crash could resurrect stale.
• Report streaming — rows come from several queries, so a plan republished mid-download must be detected and declared rather than producing a valid-looking hybrid.
• NUL bytes in content — persistence path was fixed once already; check it holds.

Verification

go build ./..., go vet ./... clean. server/store suite green.