Pass on g_round_even = true default, to support using false for golden reference results in tb_fil_ppf_wide_file_data.vhd.

Use g_round_even = false in tb, because golden results use round half away instead of round half even.

Use applications/lofar2/model/pfs_coeff_final.m to create FIR coefficients for 16b or 18b Kaiser and Hanning.

This reverts commit 2b13fea1

Removed g_r2_mul_extra_w because it more clear to increase stage_dat_w. Removed g_sepa_extra_w because separate rounding is no moved to FFT output requantizer.

Round outside separate function in output quantizer, to avoid more inaccurate rounding of 1 LSbit in separate for bit growth.

This reverts commit 37dfdaea