chain 0.23.x: Fix ChainPosition ordering

[evanlinjin]

Description

Back-ported from #2074
Depends on #2148

This PR fixes incorrect ordering of ChainPosition that affected how wallet transactions are displayed. Previously, unconfirmed transactions that were never seen in the mempool would incorrectly appear before transactions with mempool timestamps due to the derived Ord implementation treating None as less than Some(_).

Problem

The derived Ord implementation for ChainPosition::Unconfirmed caused incorrect transaction ordering:

• Unconfirmed { first_seen: None, last_seen: None } (never in mempool)
• Would appear before Unconfirmed { first_seen: Some(10), last_seen: Some(10) } (seen in mempool)

This resulted in a confusing wallet display where transactions never broadcast would appear before pending mempool transactions.

Solution

Implemented manual Ord and PartialOrd traits for ChainPosition with the correct ordering:

1. Confirmed transactions (earliest first by block height)
2. Unconfirmed transactions seen in mempool (ordered by first_seen)
3. Unconfirmed transactions never seen (these come last)

Additional ordering rules:

• At the same confirmation height, transitive confirmations come before direct confirmations (as they may actually be confirmed earlier)
• Tie-breaking uses transaction IDs for deterministic ordering

Changelog notice

### Fixed
- Fixed `ChainPosition` ordering so unconfirmed transactions never seen in mempool appear last instead of first

### Changed
- `ChainPosition`, `CanonicalTx`, and `FullTxOut` now require `A: Ord` for their `Ord` implementations
- Simplified `FullTxOut` ordering to only use essential fields (chain_position, outpoint, spent_by)

Checklists

All Submissions:

• I followed the contribution guidelines

Bugfixes:

• This pull request breaks the existing API
• I've added tests to reproduce the issue which are now passing
• I'm linking the issue being fixed by this PR

[luisschwab]

ACK c3e960f

[notmandatory]

ACK c3e960f

I ran it through the LLM 🤖 reviewer also and its only comments were about missing edge case tests with invalid ChainPosition data. But preventing invalid data should already be covered in other tests.