server: Always serve known getcfilterv2 filters.

[davecgh]

Similar to the recently changed getheaders handling, which also had the same semantics, the current logic for responding to getcfilterv2 includes a check to avoid serving filters whenever the current local chain is not considered to be current.

Like the getheaders case, one less than ideal consequence of not responding is that it can lead to peers appearing to be unresponsive and/or stalled whenever the peer does not consider itself current, which can happen temporarily in a variety of corner cases such as after being unable to communicate with the network for a long time, or in testing scenarios where there are necessarily long periods of time without any new blocks.

This existing behavior for getcfilterv2 handling was inherited from the older bloom filter logic where it mattered because bloom filters could potentially put a substantial strain on the server and also could significantly slow down the initial chain sync.

However, the aforementioned concerns no longer apply to version 2 block filters because they have a fixed creation cost per block, are required to exist for main chain blocks by consensus, and are the same for all peers.

Taken as a whole, this means that serving the filters is relatively cheap and therefore this modifies the semantics to always respond to getcfilterv2 requests, even before the local chain is fully synced.

[matheusd]

This now works to serving cfilters to the wallet once the minKnownWork height is reached, as expected.