Investigate sign flipping of ric(nohc) in regression tests

The ric(nohc) variable is currently ignored in regression tests (on the 1092 python conversion branch) because in the i_mode scenario the sign of the value is consistently different between local and CI test runs of this scenario. This is documented further in #1118.

This seems to be due to a comparison of two identical values, where the rounding error on them can result in a sign flip of the value of ric(nohc). The magnitude does not change, only the sign. The sign of the value has been seen to alternate between positive and negative at least twice during a run of the i_mode test on the CI system.

This failure seems to be a product of a bug (or at least weakness) in Process itself that appears to be longstanding, and was possibly masked in the old test suite due to a problem detecting negative diffs, although I am not certain of that.

Ideally an explanation for the difference between the local (i.e. my machine) and CI system results for the value of ric(nohc) needs to be found. My theory of "random" rounding errors (which consistently produce opposing results) does not sit well with me. If this can be explained, then a better solution for the sign of this value should be found, perhaps by using a rounding-error tolerant comparison.