patch: Parquet Column Names with "Special Characters" fix

[MarquisC]

We're using PyIceberg to read Iceberg tables stored in S3 as parquet. We have column names in the form of id:foo diagnostic:bar using : as a sort of delimiter to help us do some programatic maintenance on our side.

In Parquet the column names are magically subbed in this case : -> _x3A and upon attempts at scanning/reading the data the schema of the table doesn't match the physical column names for PyArrow.

The first pass is a naive fix for this that I have tested and works, but I'm looking for guidance on where you all want me to put this logic, and I'm happy to add it there instead.

[mchamberlain-mdsol]

Exception Example (before hack):

No match for FieldRef.Name(prefix:foo) in prefix_x3Afoo: string not null
... a bunch of other columns 
__fragment_index: int32
__batch_index: int32
__last_in_fragment: bool
__filename: string

After the hack I can read the dataframe, would love some guidance on where you all think something like this is most appropiate.

[Fokko]

Thanks for raising this @MarquisC. This looks like #83, can you check if that also resolves your problem? Otherwise, I think it will be a good place to add it here as well. It also shows how to test this.

[MarquisC]

Will test and close, unit test wise looks like #83 is the fix.

[mchamberlain-mdsol]

Will test and close, unit test wise looks like #83 is the fix.

This PR can be closed, confirmed fix is in master.

[Fokko]

Thanks you both @MarquisC and @mchamberlain-mdsol for checking this!