add optimization for regression jobs in MRs with no parallel float branch

    ### ----- run pytest for merge target now -----

    # create float reference outputs with corresponding branch

    - python3 -m pytest $TEST_SUITE --update_ref 1 $enc_stats_arg $enc_dmx_arg --create_ref -n auto --ref_encoder_path $REF_ENCODER_PATH --ref_decoder_path $REF_DECODER_PATH || exit_code=$?

    # optimization: for MRs with no parallel branch in float, REF_[ENCODER|DECODER]_PATH and MERGE_SOURCE_FLOAT_REF_[ENCODER|DECODER]_PATH are the same (see wiki: https://forge.3gpp.org/rep/sa4/audio/ivas-basop/-/wikis/CI-Environment#how-are-regressions-calculated).

    #               Hence, we only need to do a second reference creation run if the paths are distinct as the result would be the same otherwise. This saves time for branches with no parallel branch in float.

    - |

      if [ $REF_ENCODER_PATH != $MERGE_SOURCE_FLOAT_REF_ENCODER_PATH ] || [ $REF_DECODER_PATH != $MERGE_SOURCE_FLOAT_REF_DECODER_PATH ]; then

        python3 -m pytest $TEST_SUITE --update_ref 1 $enc_stats_arg $enc_dmx_arg --create_ref -n auto --ref_encoder_path $REF_ENCODER_PATH --ref_decoder_path $REF_DECODER_PATH || exit_code=$?

      fi

    - bash "${CI_PROJECT_DIR}"/ivas-codec-ci/snippets/basop/overwrite-pytest-cache-with-artifact.sh

    # create merge-target outputs

    - python3 -m pytest --tb=no -q $TEST_SUITE --keep_files --create_cut --html=$HTML_REPORT_MAIN --self-contained-html --junit-xml=$XML_REPORT_MAIN $comp_args --dut_encoder_path $MERGE_TARGET_ENCODER_PATH --dut_decoder_path $MERGE_TARGET_DECODER_PATH -n auto --testcase_timeout $testcase_timeout || true