Fix the network code handling reorgs to not cause the server to be disconnected when a block is parked and the next block takes some time to replace it.

Test how Electrum handles a reorg wrt to the server_height and block_height.
In the new test case, we have one block reorged (parked by Avalanche because of a policy violation) after it was initially accepted by the node, and we test each intermediate step as the chain grows longer again.
In the existing test, we keep testing a multi-block reorg by a longer fork. We adds the checks for the local and server height.

Depends on D14968