Miscellaneous L3 timing fixes

[PlasmaPower]

• Fix various places where L1 block numbers were called parent chain block numbers, and vice versa
• Fix minAssertionPeriod not being respected because L1 and parent chain block numbers were confused in l1_validator.go
• Fix header_reader.go assuming that chains with non-zero difficulty didn't have finality (Arbitrum chains have a difficulty of 1 and have finality)
• Fix parentChainBlockNumber and l1BlockNumber function in delayed.go
• Run the delayed sequencer at startup, in case the parent chain isn't producing new blocks but there are new delayed messages to sequence. This also matches the behavior of the sequencer coordinator sequencing delayed messages when acquiring the lockout.