feat: add relative equals that defaults to direct equals for small values

[kurtsansom]

For relative equals there is check in case the values are small.

This commit provides a relative equals that defaults to direct when the values are small. e.g. max(relative_tolerance*abs(expected), tolerance)

If there is a better way to do this please advise.

Originally I tried to keep with the naming convention, but ran into trouble with the length of function names and therefore used RelMinEqual which I really don't like.

Another thing that need some more attention is I didn't modify the added test yet to evaluate the threshold. It would be good to get feedback that this is the right approach before doing more.

[kurtsansom]

I decided to rename the function to approx to simplify the name length, and mirror the name use by pytest as there doesn't appear to be an equivalent in JUnit. I removed the relative difference calculation as providing the difference and calculated tolerance is sufficient to describe the criteria that is failing.

[tclune]

Looks like it would not be too difficult to introduce aliases, but I'd not ask you to implement it.

The intro would define a small dictionary whose keys are the annotations and the values are the Fortran procedure names. It would then intercept the usual expansion:

        parser.outputFile.write(fragment.format(match.group(1), match.group(2)))

[kurtsansom]

I made a simple example on this branch.

https://github.com/kurtsansom/pFUnit/tree/assert_alias

map:

assert_alias_map = {'approx' : 'ApproxEqual'}

in class AtAssert(Action):

        if (match.group(1).lower() in self.alias_map.keys()):
            function_name = self.alias_map[match.group(1).lower()]
        else:
            function_name = match.group(1)

[tclune]

Same request as the other. Please change to PR onto develop branch.

[kurtsansom]

I will shoot for getting to this by the end of the week.

[tclune]

@kurtsansom - do you still intend to resubmit this as a PR onto develop?

[kurtsansom]

@tclune Apologies this has fallen off my radar. It looks like I just missed fixing this by about a week.