Improve fms2_io performance

[uramirez8707]

Is your feature request related to a problem? Please describe.

Doman Reads:

• In the current version of fms2_io, the io_root pe does 1 read for each of the pes and sends it to the other ranks, which is very inefficient

Compressed Writes:

• In the current version of fms2_io, the io_root_pe gathers and writes the data one PE at a time, instead of just doing one write

Describe the solution you'd like
Doman Reads:

• The root pe should do 1 read instead of multiple. The sending of the data should be non-blocking.

Compressed Writes:

• The root pe should gather all of the data and do one write instead of multiple.

Describe alternatives you've considered
N/A

Additional context
N/A

[uramirez8707]

Fixed on main in #1259