Initial some variables that MemorySanitizer suggested

FWIW, a lot of these look like false positives. A limitation of the sanitizer is that it can only detect e.g. MPI's writes if the external library is itself built with the sanitizer. Duplicate initialisation is mostly harmless, especially with a trap value like MPI_UNDEFINED, but it's worth weighing against the risk of disguising actual uninitialised reads further down the line. Although, from 'also reported issues...', maybe you already know this?

Thanks, yes, I mainly included these so that I could run the memory sanitizer to look for deeper issues, which failed to materialise, but at least if I try again these will not come up.