ephemerides handling fixes

@GregAshton 2 fixes to ephemerides handling:

1. Fix a problem I introduced in !18 (merged) - for some reason I cannot remember (maybe just "high numbers look good") I picked DE421 as the standard to look for, while in lalapps it's currently actually DE405 for both CFSv2 and MFDv5.

2. Teach Writer.run_makefakedata to actually pass through the chosen ephem files to the MFDv5 command line. The search class already properly handles them through InitBarycenter, while for making SFTs it seems we never actually did this properly so far.