Parallelize Reading of Trace Files

[DavePearce]

Reading large trace files has become a fairly large bottleneck, and this is something that could potentially be parallelised (as suggested by @AlexandreBelling). The goal of this issue is twofold:

1. Check that indeed parsing bytes into fr.Element is the real bottleneck (i.e. not reading bytes from disk into memory)
2. Actually implement the parallelisation

[DavePearce]

Also to do:

1. Special case for reading of columns with smaller widths.