fix(state): Avoid panics and history tree consensus database concurrency bugs

[teor2345]

Scheduling

This PR should probably get in before the 1.3.0 release, because it fixes a panic bug.

Motivation

1. We've been seeing "missing sprout tree" panics in CI and on local machines (but users haven't reported them yet)
2. A similar concurrency bug exists in history trees, but it causes invalid empty history tree data to be returned from the state

Close #7581

Zebra Designs

These are the only column families that:

1. Delete the old height and insert the new height
2. Don't have code that handles the height changing in the middle of reading the height & data

https://github.com/ZcashFoundation/zebra/blob/main/book/src/dev/state-db-upgrades.md#current-state-database-format

(The UTXOs get deleted, but missing UTXOs are handled correctly, and their keys are never updated.)

Complex Code or Requirements

This fix to the tree reading code is compatible with the permanent fix in PR #7392: changing to the key format to an empty key, so the trees get overwritten rather than deleted and re-inserted.

Solution

• Always read the last sprout tree, regardless of the key value or format
• Always read the last history tree, regardless of the key value or format
• Add a RawData database serialization type

Review

This PR should probably get in before the 1.3.0 release, because it fixes a panic bug.

Reviewer Checklist

• Will the PR name make sense to users?
  • Does it need extra CHANGELOG info? (new features, breaking changes, large changes)
• Are the PR labels correct?
• Does the code do what the ticket and PR says?
  • Does it change concurrent code, unsafe code, or consensus rules?
• How do you know it works? Does it have tests?

Follow Up Work

Permanent format fix in PR #7392.

[arya2]

Looks good!