Refactor infinity substitution

[lxvm]

Checklist

• Appropriate tests were added
• Any code changes were done in a way that does not break public API
• All documentation related to code changes were updated
• The new code follows the
  contributor guidelines, in particular the SciML Style Guide and
  COLPRAC.
• Any new documentation only uses public API

Additional context

Fixes #141
Addresses a sub-issue of #221

This pr refactors the infinity substitution and expands the test set to include checks for:

• batch integral functions
• inference succeeding through the infinity transformation
• flipped limits of integration

To improve type-stability, we could always do the inf transformation by default, which could address #65. I haven't done that here since it is a major change that is most likely non-breaking, but I want to bring it up for discussion.
Additionally, I think handled units more correctly, but I don't have tests for that yet.

[ChrisRackauckas]

To improve type-stability, we could always do the inf transformation by default, which could address #65. I haven't done that here since it is a major change that is most likely non-breaking, but I want to bring it up for discussion.

I think we should. It should be an optimization to do transform = Val(false) (with type information), so that things always work by default. I don't think the cost of doing a runtime branch check is high, so just having a runtime "if transform" is really way too cheap to be paying the cost of type instability.

[araujoms]

I just wanted to thank you for this PR; I had a bug I was meaning to fix for a while (caused by mismatched types when using Inf boundaries), and finally got around to it today. But to my surprise the buggy piece of software mysteriously started working. I then tracked what changed, and it turned out that version 4.4.0 of Integrals is what fixed my bug, and it was released just yesterday!