Update SBA CLDFB-Domain HRTFs

During the XC of #399 (closed) we observed large differences between the CLDFB-domain converted HRTFs and the time-domain HRIRs, but decided to accept the ticket in order to close contribution 036 and unblock progress towards selection testing. Since then we have continued investigating the conversion from time to CLDFB-domain and include here an updated set of SBA CLDFB-domain HRTFs which closely match the time-domain HRIRs whilst also reducing the number of taps from 5 to 3. In addition, the updated set reduces the renderer delay by 31 samples. Improvement is observed in PEAQ scores with the new set when compared to the baseline, now bringing the scores very close to the CREND reference.