add per-module bandwidth calculations to pydarshan job summary reports

[shanedsnyder]

(From discussion in #549)

We should add some per-module bandwidth calculations to pydarshan job summary so that users can quickly see roughly what I/O bandwidth they are achieving via different APIs. darshan-parser has some code to provide a few different bandwidth calculations that we should figure out how to expose to pydarshan. At the very least, an 'I/O bandwidth by average' and 'I/O bandwidth by slowest' would be useful to display.

[shanedsnyder]

In our current job summary tool (i.e., not pydarshan-based), we include some high-level performance information near the top of the report that gives info on total bytes moved and obtained bandwidth across a couple of modules:

I was imagining we could include some general performance numbers like this in these per-module sections we're adding to the pydarshan reports.

This is probably a nice task we could use to think about how to best share some of the performance calculations we've done in our C code to pydarshan, which we've talked off/on about so far but have thus far avoided. As far as priorities go, I think it probably makes sense to think more about once we've got the v1 job summary milestone done, though.

[carns]

I like the idea of moving this to a per-module thing.

[nawtrey]

In today's meeting we agreed that this can be held off for now since we don't have the cumulative data available on the pydarshan side of things. So probably a pydarshan-v2 thing.

[carns]

Depends on #642.

[shanedsnyder]

This was completed via #898.