Output information about time taken for program run

[alexdewar]

This was discussed in the context of writing it to the TOML output file (see #526), but I actually think this info more naturally belongs in the program logs, which we're saving to the output folder in any case.

In addition to logging the total time taken for the simulation run, we could also record more granular information, i.e. how much time was spent in different steps of the algorithm (e.g. dispatch optimisation, agent investment). I'm in two minds about this part, as I don't want us to fall into the trap of trying to write our own profiler (which will never be as good as a proper tool), but we could provide this information at a really coarse level if users are keen.

[ahawkes]

Yeah, not sure either. I'd like this info during development for sure. Also I can imagine sets of input data that results in really slow optimisation and/or really slow agent investment. So some basic profiling to users could be useful to them - and aid in making better input data.

[dalonsoa]

I'm starting to work on this, and in order to target the right level of granularity, it will be useful to know exactly what functionality is meant to be timed.

If we are interested in high level information, I would suggest to time things in the run function, in particular:

• perform_agent_investment
• The multiple calls of run_dispatch_for_year (this, this and this)

But I don't know if you want to drill to lower level of detail. Anywhere else?

Also, should this time information - in principle included in the log file - be optional or always there. Being so high level, it should not have a big impact performance-wise, so we can very well include it always. But up to you.

[dalonsoa]

I've created a PR with some tentative, very simple implementation, #1421 . Please, have a look when you have time.