Instead of comparing the blockhash against the emitted_blocks map
to see whether the block is part of the emitter's best chain, we
reduce the `last_mempool_tip` height to the last agreement height
during the polling logic.
The benefits of this is we have tighter bounds for avoiding re-
emission. Also, it will be easier to replace `emitted_blocks` to
a `CheckPoint` (since we no longer rely on map lookup).