Update WaitForBlock to maintain two timeouts: a no-change timeout, which fires if the chain's head does not change within a specified window, and an absolute timeout, which fires if the chain's head does not meet or exceed the specified block.

These changes should ideally reduce the number of test flakes we're seeing. Everything takes longer when test executors are under load; by maintaining these two timeouts we can provide longer-running tests with more buffer while retaining the ability to fail fast if the chain gets stuck.

As part of this PR I also refactored the wait method to use polling rather than WebSockets. I've found WebSockets to be unreliable in tests.