Advanced resistant TX finality

[ggwpez]

This umbrella issue is for collecting ways to improve our TX finality (#97):

• ContractBackend.confirmNTimes cannot deal with the case that a reorg occurs with less blocks but more combined TD (total difficulty) exactly between its head and receipt polling logic. This case is not testable with the sim backend from geth since the difficulty per block is fixed. (from Confirm TX with Finality Depth go-perun#101 (comment))
• Re-sending TX that were dropped from the mem pool
• Gas bumping slow/dropped TX
• Resistant header sub