This PR adds a job which runs once a minute that prunes unnecessary txn sequences from the corpus. It removes sequences that no longer contribute any new coverage compared to the rest. This can happen when, for example, sequence A adds new coverage, sequence B builds upon sequence A and adds strictly more coverage, making sequence A now unnecessary since it doesn't "contribute anything" to the corpus.
The job works by:

• Shuffle the corpus
• Initialize a blank coverage map
• Run through the corpus one-by-one. For each sequence, run the sequence and add coverage to the coverage map. If there is no new coverage, we can remove this sequence from the corpus

The job takes about 5-10 seconds, during which time no fuzzing at all can occur

This seems to be pretty effective at reducing corpus size. In a small test I've been running, it's been removing about 100 entries per minute during the initial phase of fuzzing where the corpus quickly grows to 1000 entries (first 5 minutes) and has now slowed down to removing about 50 entries per minute (15 minutes in, corpus size 1400). Corpus growth is slow at this phase because of pruning; the corpus would probably have 2000-2500 entries by now (rather than 1400) if not for pruning.