feat: build, chain-validate, and seed evo snapshot state

[PastaPastaPasta]

Issue being fixed or feature implemented

Stacked on #7592 — the first two commits shown here belong to that PR; this one adds the last three. It will be rebased once #7592 merges.

Part of the AssumeUTXO M4 decomposition (#7579 — see the series map there); stacked on the format-core PR and reviewable as one question: given the v3 format, how does snapshot content get produced from chain state, verified against it, and seeded back into EvoDB? The dumptxoutset/load lifecycle wiring is deliberately absent — it is the next and final PR of the series.

What was done?

Three layers, one commit each:

1. Build & chain-validate (evo/snapshot.cpp): BuildEvoSnapshot() collects active/safety commitments, rotation snapshots, historical MN-list diffs, and exact score modifiers from chain state; ValidateEvoSnapshotAgainstChain() verifies a decoded snapshot against the block index, deployments, work-block ancestry, rotation horizons, and per-entry seeded modifiers.
2. EvoDB seeding surface (deterministicmns, llmq/blockprocessor, llmq/snapshot, creditpool, mnhftx, evodb): Seed* writers that store exactly the state reconstruction later reads, without publishing into the shared NORMAL-chainstate caches. GetHashModifier() prefers a seeded exact modifier when the work block's data is unavailable and cross-checks it against recomputation when it is; a mismatch throws SnapshotStateMismatchError, which is unreachable in production until the load integration seeds modifiers (the next PR routes it into the controlled invalid-snapshot path — the plumbing is deliberately not wired here). GetDeterministicMNListHash() switches to the canonical codec hash so completion markers and payload-derived hashes share one identity.
3. Partial-history support: BuildEvoSnapshot() originally required every enabled LLMQ type to carry exactly signingActiveQuorumCount actives plus full safety and rotation horizons, which a young chain, a freshly activated type, or a rotation type with pre-activation horizon cycles cannot satisfy — dumptxoutset would fail on valid chain state (flagged as blocking by review on backport: assumeutxo M4 — evo snapshot format v3 and LLMQ reconstruction #7579). The builder now emits the history that exists, context-free validation treats parameter counts as maxima, and chain validation requires every carried entry to sit at a derived horizon position while tolerating absent ones. This does not make withholding easier: the completion-time CbTx quorum merkle root pins the active set exactly, historical-diff and modifier tallies must mirror the carried commitments, and completion-time quorum reconstruction establishes sufficiency. Two regression tests (young-chain builder, partial-history validation) fail against the previous exact-count enforcement.

One accepted dev-channel caveat: the GetDeterministicMNListHash() encoding switch means a develop node that activated a snapshot before this change and completes validation after it would fail completion conservatively (marker hash mismatch) and re-sync from the base. No released version carries the old markers.

How Has This Been Tested?

Full unit suite green on a --enable-werror build; the chain-aware test block exercises reconstruction from seeded state only, pre/post-DIP3 validation matrices, rotation bitset/work-list binding, and the two new partial-history regressions. Snapshot suites also run under a --with-sanitizers=undefined,integer build.

Breaking Changes

None on any release branch (see the dev-channel caveat above).

Checklist:

• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have added or updated relevant unit/integration/functional/e2e tests
• I have made corresponding changes to the documentation
• I have assigned this pull request to a milestone (for repository code-owners and collaborators only)