Regex Bugfix

[GuyARoss]

We encountered an issue when utilizing this package as an informer sync mechanism where snake_case fields were not correctly being cached.

I traced the issue to this regex, modified it to accept snake_case, which seemed to fix the issue.

[crobby]

Thanks for your PR. It looks like it's breaking our current set of tests, which would need to be fixed.
Additionally, you would need to include new tests that cover the cases for which the regex is being modified.

[ericpromislow]

The failed test is sometimes a flake, but this PR iguarantees failure. The failing test relies on a setting for a field called "_type". The passing test maps "_type" to "type" and succeeds in finding values in obj.type but not obj._type.

There's some object transformation going on as well that maps fields called "type" to "_type" so we can bypass metaprogramming that does other things on an object's type (sorry).

The test passes if you pull the _s out of the modified regex:

subfieldRegex = regexp.MustCompile(`([a-zA-Z0-9]+)|(\[[-a-zA-Z0-9./]+])|(\[[0-9]+])`)

But this doesn't help you at all ... you'll need to do some more work, possibly looking at the transforms
based at pkg/resources/virtual, to figure out what needs to be done.

[ericpromislow]

Unit tests fail. More work is needed to support field names containing underbars.