gracefully handle bogus .available.shards files

[seebs]

What's going wrong?

If someone's filesystem creates 0-byte .available.shards files, pilosa can fail to start up because it can't read them.

What was expected?

Well, probably a working filesystem, but I think that's out of scope. I think we should handle that case gracefully.

So far as I can tell, the .available.shards files are functionally a cache -- they can be recreated at runtime by querying other nodes. If we can't read them, we should ignore them rather than failing.

Steps to reproduce the behavior

Zero out the files while pilosa is down, start it up.

Information about your environment (OS/architecture, CPU, RAM, cluster/solo, configuration, etc.)

Affected systems were Centos using XFS, but it's irrelevant.

[travisturner]

@seebs has any work been done on this? should we keep it as "In Progress"?

[seebs]

It's actually the first commit in my fsck branch (which means it's stacked under a ton of other changes), so a fix exists, it's just not merged.