Fix the GEC for FT version > 4

This MR fixes the bug in setting the GEC cut for FT decoding version >4 that was originally spotted in the Moore tests (see Moore!325 (merged).

Although many files potentially had the bug, only one test is actually affected:

more build.x86_64-centos7-gcc8-opt/html/Brunel.tmph4ck.Full_HLT1_FTv6_MDF/output_diff
<pre>- LAZY_AND: moore  #=1000    Sum=966         Eff=|( 96.60000 +- 0.573097)%|
?                                 ^^                 ^ ^           ^ ---
+ LAZY_AND: moore  #=1000    Sum=935         Eff=|( 93.50000 +- 0.779583)%|
?                                 ^^                 ^ ^          +++ ^
-  PrGECFilter     #=1000    Sum=970         Eff=|( 97.00000 +- 0.539444)%|
?                                 ^^                 ^     -       ^ ---
+  PrGECFilter     #=1000    Sum=939         Eff=|( 93.90000 +- 0.756829)%|
?                                 ^^                 ^ +          + ^^^
-  FilterOnNoPVs   #=970     Sum=966         Eff=|( 99.58763 +- 0.205760)%|
?                     ^^          ^^                    - ^^       ^ ^^^
+  FilterOnNoPVs   #=939     Sum=935         Eff=|( 99.57401 +- 0.212538)%|
?                     ^^          ^^                     ^^^       ^^ ^^
-  UpgradeForward  #=966     Sum=966         Eff=|( 100.0000 +- 0.00000 )%|
?                     ^^          ^^
+  UpgradeForward  #=935     Sum=935         Eff=|( 100.0000 +- 0.00000 )%|
?                     ^^          ^^
-  VeloKalman      #=966     Sum=966         Eff=|( 100.0000 +- 0.00000 )%|
?                     ^^          ^^
+  VeloKalman      #=935     Sum=935         Eff=|( 100.0000 +- 0.00000 )%|
?                     ^^          ^^

Since the tmph4ck tests are obsolete and are being replaced by equivalent tests in Moore (see Moore#34 (closed)), I leave it to the Brunel maintainer whether to merge this MR (will require updating Rec/Brunel/tests/refs/Full_HLT1_FTv6_MDF.ref) or just wait for these tests to be removed. @pkoppenb @mstahl FYI