dwi2fod: Catch underdeterminedness

[Lestropie]

Have caught a few users falling for this mistake, and would prefer to not have to doubt it every time.

Currently, if the user requests more ODFs than there are b-value shells, the command just proceeds regardless. Typically this involves requesting the common three-tissue decomposition for single-shell data yielding an empty GM ODF. People who don't perform adequate data checking may completely overlook this. dwi2fod msmt_csd should be aborting up-front if such an attempt is made.