Parse error handling

[gwillen]

This is not really ideal, but I think it's an improvement over the current behavior: Instead of arbitrarily and silently stopping in the middle of a keyring when we fail to parse a packet, we skip the failed packet and continue.

There is a remaining unfortunate issue, which is that the implied shape of the resulting keyring depends strictly on all appropriate packets being present; if we drop a certificate, e.g., but not the subsequent UIDs, they will appear to be on the wrong certificate.

On the other hand, I am still trying to track down at least one case where the current code appears to already have this problem (dropping a User Attribute packet containing a JPEG, but retaining the following signatures which appear to be on the preceding UID packet.)

Up to you whether you want to pull this or wait for me to figure out how to improve upon it (probably by allowing us to create placeholder objects for Certificates we can't parse.)