BUG: Fix large floats in Excel losing precision when converted to integer

[ng-henry]

• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.

When opening Excel files with large floating point values like 1E50, Pandas will convert these values to integer, resulting in integers like 100000000000000007629769841091887003294964970946560. Converting large floating points to integers results in erroneous values.

This PR only converts floats to integers if they are smaller than 1E22, which is about the maximum float that can fully represent an integer. If any floats are higher than that cutoff, then they are not converted to integers.

[debnathshoham]

Thanks for the PR @ng-henry !
Is there a related bug report? If not, could you please open one

[mroeschke]

Thanks but these constants like 1e22 are not very robust to platform difference and can hide bugs. Unless there's a openpyxl flag or package constant that can be used, I recommend just documenting this limitation.

[ng-henry]

Since we know Python represents floats as 64 bit doubles, we are guaranteed that all doubles have 16 decimal digits of precision (according to https://en.wikipedia.org/wiki/Double-precision_floating-point_format). So I'm thinking we can replace 1e22 with 1e16? 1e22 was just an empirically determined value, but 1e16 is based off the maximum precision of a double number.

On a practical note, numbers larger than 1e16 are probably better represented with floats than integers. If an Excel cell contains numbers that large, the meaning is better captured as a float than as an integer.

For context, this int(cell.value) check was done because of #46988.

[mroeschke]

Thanks for the pull request, but given the issue and solution, I think more discussion is needed in a dedicate issue first. Closing this for now, but happy to reopen if other core developers think this is an adequate solution in the issue