- explicitly setting inchannels and outchannels is now optional,
  and default values are pulled from the input stream argument
  of AdjustedMatrix.process() method
- *if* non-F inchannels and outchannels are specified, they are
  checked for dimensional consistency with each other, and with
  self.matrix
- returns dimensionally consistent array of NaNs if there is any
  mismatch in inchannels, outchannels, or self.matrix
- raises exception only if all inchannels are not provided in
  the input stream
- requires F be specified in both inchannels and outchannels, or
  else it returns NaNs for F

This sounds a little complicated, and it is, but it was necessary
in order to allow the AdjustedMatrix class to adjust both vector
values (e.g., HEZ->XYZ), as well as scalar F values (e.g., pier
corrections). An argument could be made to separate these, but I
am not willing to do so at the moment.