Check for extreme values in skims and log error

[bricegnichols]

In export skim process in SkimsAndPaths (before averaging) check for configurable max values to be set in network_configuration.toml and report as a warning in a log for skims. Include set of ODs when possible, maybe max out at first 50?

[bricegnichols]

Make changes on asim_develop branch - can run with activitysim or daysim for development, should work with either.

In SkimsAndPaths, the function average_skims_to_hdf5_concurrent exports skim matrices from Emme to h5 files. While the matrices are in memory we should check if any of the values are very high or NaN. This can cause issues in Daysim/Activitysim later.

When the matrix_value array is available, it should be passed to a new function that checks for extreme values. This should happen before skims are averaged. (We average current skims with those from previous assignment and skimming iterations to improve model convergence).

The function should check all values are beneath some threshold value set in network_configuration. You can use a smaller value for testing like 120. The values may need to vary by the skim type (different numbers for distance vs cost), but we can add that later. In general the major issues come up when we get values like 1e9.

If extreme values are found the matrix name and time period should be added to a log file. Ideally we would report some of the O-D pairs, up to a max of 50 (?) to spot check.