An attempt to fix ssec/polar2grid#691

Basically, current bounding operations via get_area_slices have trouble when the resulting lon/lat extents match almost exactly with the source area definition. This PR should fix a couple important bugs in the array index retrieval methods of the AreaDefinition:

1. Floating point precision issues where a X coordinate that is just outside the area (a float precision different) can still be considered part of the area (not masked).
2. Not including -0.5 to 0.0 pixel indexes as "0" (it is the left most part of the left most pixels.
3. Including an extra 0.5 pixels on the right of the area.

Locally I see one failure where my particular approach has changed the underlying value for elements that are masked and this test specifically checks that that doesn't happen (they stay unchanged).

• Closes #xxxx
• Tests added
• Tests passed
• Passes git diff origin/main **/*py | flake8 --diff
• Fully documented