WP_MOK hardcoded to 8 bytes, breaks single-precision builds

[sbryngelson]

Bug

Multiple files compute `WP_MOK = int(8._wp, MPI_OFFSET_KIND)`, hardcoding 8 bytes as the working precision real size. When MFC is built in single precision (`wp = real32`), reals are 4 bytes, but MPI I/O operations use an 8-byte offset per element.

Files: `src/post_process/m_data_input.f90`, `src/pre_process/m_data_output.fpp`, `src/pre_process/m_start_up.fpp`, `src/simulation/m_data_output.fpp`, `src/simulation/m_start_up.fpp` (10 occurrences total)

Impact: MPI parallel I/O reads/writes the wrong number of bytes per element in single-precision builds, corrupting all file I/O.

Severity: High — all parallel I/O is broken in single-precision builds.

[sbryngelson]

Fix: #1187