close: Do not attempt a unilateral close when we see another close on-chain

[cdecker]

In Greenlight we have an issue with channels that close, and the
signer got desynced from the Node due to a non-atomic update which
bumped the VLS signer commitment number, but the corresponding bump
got lost on CLN due to preemption. The problem here is that the signer
will not sign off on any future states, so the channel is unusable and
needs to be closed. If the peer closes the channel however, the last-
ditch attempt of publishing the last state, i.e., signing the n-1th
commitment TX, when VLS believes we are at the nth, won't work.

To get past this situation we selectively skip the unilateral close
attempt on our end, by setting passive=true when calling
drop_to_chain. This is safe because the way we got to this place is
by seeing our counterparty's unilateral close confirmed in a block we
are processing, therefore it is already too late to publish the
unilateral close anyway (it conflicts with the counterparty's close).

Changelog-Changed: close: We no longer attempt to publish a unilateral close that'd fail anyway when we witness a close onchain.

[cdecker]

I have removed the bip32_pubkey check before starting onchaind. @rustyrussell IIRC this was put in place in order to guard against the cosmic-ray-bit-flip in derived addresses we've heard stories from the LND team. I think this check could be added to onchaind such that it's the part that locks up, would you like me to re-add that check there?

[cdecker]

The passive argument was a bit weird, now it's called rebroadcast as that's closer to what we're actually trying to do.