Optionally use ClickHouse DB to log process, run, and file metadata

[joefowler]

Not sure you'll be excited to add an optional DB-logging feature, but I do believe that it's feature-complete for a certain first set of tasks. I made enough changes that I'd like to bake this into the main branch. It shouldn't have any effect on computers where DB use was not explicitly asked for in the past.

This PR does not log pulse data, but only:

1. A row of info about this current invocation of the Dastard process (Go version, git hash, process start time and eventually, its end time)
2. A row of info about each data run (intention, source type, directory path, # of channels, timebase, number of samples and presamples)
3. A row of info about each file (filename, file type, start time, and after it's closed: the end time, record count, file size, and a sha256 hash of the file).

Again, there are no pulse records stored in the DB--no LJH-like or OFF-like data. I experimented with that possibility, but the performance was not acceptable. (See branch:clickhouse_pulse_storage for that code.)

[ggggggggg]

What benefit do we gain by merging this vs just keeping it in a branch/PR until we actually want to test the features?

[joefowler]

What benefit do we gain by merging this vs just keeping it in a branch/PR until we actually want to test the features?

I am not sure. My motivation for this PR is simply that both this work and the possible work on simplifying our data writing touch/would touch the same lines of Go. If we're going to change how we write LJH and OFF files, I don't want to make incompatible changes for that purpose and for the DB writer. I'd rather launch the one branch from the other.

Of course, that doesn't require us to merge this PR yet. We can talk in person....