bdk_electrum: cache merkle proofs

[notmandatory]

"You cache transactions but not their anchor's validity, which significantly reduces the gains from caching as you need to make a request for Merkle proofs anyways."

[LagginTimes]

I'll probably take a stab at this as well and throw it onto #1675.

[LagginTimes]

I'll probably make a separate PR for this. The previous bdk_electrum rework that introduced validate_merkle_for_anchor() is reliant on transaction_get_merkle() to update our Header information in case of a reorg. Caching merkle proofs would mean that validate_merkle_for_anchor() will no longer be aware of reorgs and will may insert anchors for transactions that have been reorged out.

After discussion with Evan, our idea is that we should instead:
1.) Store the Header in CheckPoint (and use this as the cache). This way we can keep our Headers consistent throughout sync/scan instead of updating Headers inside validate_merkle_for_anchor().
2.) If we pull a fresh merkle proof and the information does not match up with its Header, we simply do not create an anchor and let the sync/scan continue.