[evlatests] L302 update

[Ken Sowinski]

Hichem found that the MIB does not always do floating point arithmetic
correctly.  At least when computing the DDS frequency command it will
sometimes get the bottom few bits wrong.  We have reformulated the
computation so that this error is not triggered, at least not in the
test cases that were examined.  With the hope that this may go some
way to explaining the 20 microHz frequency offset that has been
reported we have updated all the L302 MIBs with SW that avoids
triggering this error.  By way of explanation I note that a
somewhat non-standard technique was used to place the 48 bits
required by the DDS in the right place of a double so that they
can be easily sent out the SPI bus.  This technique was tested
on other machines and has worked flawlessly.

Reports of frequency offsets, or other L302 related problems, or
lack of these are welcome.