PredictFstat maxStartTime fix (new alternative option: --duration)

Description

• First I noticed an omission in the test script: since the --maxStartTime option expects the starting timestamp of the last SFT to cover, not the end time, this test script was actually calling PFS the wrong way. This slipped through because of an over-generous 1% tolerance - actually the 3 runs which all use assumeSqrtSX and just different ways of selecting timestamps should be exactly equivalent, and indeed they are after the fix; but I've now set it to 1e-6 just to still be safe against cosmic rays. ;)

• Then, as described in #370 (closed), I realised that this behaviour of --maxStartTime when no --DataFiles are given is actually inconsistent with the opposite case and the underlying half-open-interval policy defined by XLALCWGPSinRange. So this extended MR now introduces a new --duration option to be used when when no --DataFiles are given, which should behave consistently with both human expectations and the underlying convention, while disallowing --maxStartTime in this situation so that no old commandlines silently change behaviour.

API Changes and Justification

Backwards Compatible Changes

As usual with lalapps changes, these are not technically API changes. A new commandline option is added and some previously supported combinations of options now throw errors, though.

• This change introduces no API changes
• This change adds new API calls

Backwards Incompatible Changes

• This change modifies an existing API
• This change removes an existing API

If any of the Backwards Incompatible check boxes are ticked please provide a justification why this change is necessary and why it needs to be done in a backwards incompatible way.

Review Status

To demonstrate the first (purely test script) fix, the relevant part of the test log after tightening the tolerances, but still using endTime:

SemiAnalyticF:              2F_SA  = 51093.8
PredictFstat{assumeSqrtSX}: 2F_PF0 = 51093.7
PredictFstat{NoiseWeights}: 2F_PF1 = 47034.4
PredictFstat{timestamps,assumeSqrtSX}: 2F_PF2 = 51093.7
PredictFstat{minStartTime+maxStartTime,assumeSqrtSX}: 2F_PF2 = 51283.8

Relative deviation 2F_PF{assumeSqrtSX} wrt 2F_SA = 0.0002% (tolerance = 1%) ==> OK.
Relative deviation 2F_PF{NoiseWeights} wrt 2F_PF{assumeSqrtSX} = 8.2735% (tolerance = 20%) ==> OK.
Relative deviation 2F_PF{timestamps,assumeSqrtSX} wrt 2F_PF{assumeSqrtSX} = 0.0000% (tolerance = 0.0001%) ==> OK.
Relative deviation 2F_PF{minStartTime+maxStartTime,assumeSqrtSX} wrt 2F_PF{assumeSqrtSX} = 0.3714% (tolerance = 0.0001%) ==> FAILED.

after correctly using maxStartTime:

SemiAnalyticF:              2F_SA  = 51093.8
PredictFstat{assumeSqrtSX}: 2F_PF0 = 51093.7
PredictFstat{NoiseWeights}: 2F_PF1 = 47034.4
PredictFstat{timestamps,assumeSqrtSX}: 2F_PF2 = 51093.7
PredictFstat{minStartTime+maxStartTime,assumeSqrtSX}: 2F_PF2 = 51093.7

Relative deviation 2F_PF{assumeSqrtSX} wrt 2F_SA = 0.0002% (tolerance = 1%) ==> OK.
Relative deviation 2F_PF{NoiseWeights} wrt 2F_PF{assumeSqrtSX} = 8.2735% (tolerance = 20%) ==> OK.
Relative deviation 2F_PF{timestamps,assumeSqrtSX} wrt 2F_PF{assumeSqrtSX} = 0.0000% (tolerance = 0.0001%) ==> OK.
Relative deviation 2F_PF{minStartTime+maxStartTime,assumeSqrtSX} wrt 2F_PF{assumeSqrtSX} = 0.0000% (tolerance = 0.0001%) ==> OK.

To demonstrate the fix of #370 (closed), repeating the example commandlines from over there will now produce an error from the last two, but changing them to use --duration instead now produces

$ lalapps_PredictFstat --outputFstat=$tmpPFS --h0=1 --cosi=-0.3 --psi=0.6 --Alpha=2 --Delta=-0.5 --assumeSqrtSX=0.5 --IFOs=H1 --minStartTime=700000000 --duration=1800 && grep Loaded $tmpPFS

228.4
%% Loaded SFTs: [ LHO_4k:1 ]

$ lalapps_PredictFstat --outputFstat=$tmpPFS --h0=1 --cosi=-0.3 --psi=0.6 --Alpha=2 --Delta=-0.5 --assumeSqrtSX=0.5 --IFOs=H1 --minStartTime=700000000 --duration=3600 && grep Loaded $tmpPFS

421.9
%% Loaded SFTs: [ LHO_4k:2 ]

which is consistent with the other results.