i97 Insert observations grouped first by network

[QSparks]

Added two functions to insert.py that will sort an arbitrary list of observations into a dictionary where <network_name> is the key. This way network-level logging is maintained even if calls to insert have observations from multiple networks.

resolves #97

[jameshiebert]

The overall structure of the approach that you've taken looks great, but I have a couple of questions/comments that I think need to be addressed.

If you're able to write some unit tests that exercise these new functions, I think that would be helpful, because my read of the get_network() function is that it will not work (w.r.t. the caching) as written.

[jameshiebert]

It's looking much better! Good work 👍

A few more comments that I think will further improve the code.

[jameshiebert]

Looks much better! Last two comments and then I think we're done:

1. I think log_helpers as a separate module is a good idea for these two functions. However, the name is a bit of a misnomer, since cached_function isn't really used in logging. How about db_helpers.py?
2. I think that it's important for db_metrics to be returned (or yielded) per network. Different networks have different performance characteristics. I'd like to you swap the loops there and have each per-network iteration yield the db metrics results. You may need to adjust the calling code in process.py to make it work.