feature_fedpeg.py: just use testmempoolaccept checking blinded peg-in…

[instagibbs]

… transactions

Previous test had race condition causing spurious failures for no additional coverage.

Resolves #728

[instagibbs]

ok, somehow sidechain2 sees the non-pegin input as already spent, much like the previous test setup. The added mempool check shows that it's unrelated to the later disconnect and block creation.

[instagibbs]

ok it's the obvious thing, a block race when selecting this additional input. Previously the test raced forward and could fail because sidechain2 hasn't seen the sourced input.

[instagibbs]

now getting Block sync timed out on the build. Um.

[instagibbs]

Probably the case where the two nodes think they have nothing of use to tell each other when sidechain spams sidechain2 with a bunch of just-immature-to-sidechain2 blocks, and they're all marked invalid.

To unstick you just need to see a new block(which means complicating the already-existing sync calls). I sprinkled more syncs during block creation instead to see if that fixes it. Another tactic would be to sync blocks before each claim, make the claim, then sync mempool only, then have sidechain2 make the block. This would ensure no peg-in-invalid blocks get buried in chains of seemingly invalid ones.

[stevenroose]

ACK 3e912de