Version packet

[chjj]

@indutny, thoughts on this? Maybe we could quicker switch to the filterload stuff by sending relay in the version packet? Does it matter? Also, the startHeight...is that even applied if we're using a bloom filter? Anyway, it looks like the framer method may have been doing a writeU32 on an undefined value (luckily bitwise ops cast undefineds to zero).

Also, right now when you call peer.watch(), bcoin will throw a nasty error. couldn't calling peer.watch() without a hash be useful just to send filterload to (potentially new) peers again?

[indutny]

Regarding relay: it should be fine to allow setting it, but for filterload purposes it is currently set to the right value, otherwise you would seen a lot of TXs and Blocks arriving before the filterload was issued.

I don't really care sending the height, but please provide a test for it :)

And good that you improved .watch() method.

[chjj]

My satoshi protocol branch could give new meaning to the relay option. Maybe instead of the pool.options.satoshi: true we could use pool.options.relay: false (with the default being true).

[chjj]

Added a test and cleaned up some small things. Should be mergeable. If it's not I can rebase + force push.

[chjj]

There was a small conflict. Rebased.

[indutny]

Thanks!