Fix remaining sec inconsistency regarding the X-IF addition

• I reviewed #290 and found out that what the main issue still missing was that the signal rf_wdata_sel, coming out of the cve2_decoder, had its length changed from 1 to 2 bits in order to allow the RF_WD_COPROC value.

The problem is that this was made independently if the X-IF interface is enabled or not, so I set it to be 1 bit when XInterface is 0.

• Regarding the reset on the rf_we, as done on the original core. I just simplified the logic on #290
• Additionally, I modified the assertion that checks the value of rf_wdata_sel after instr_valid_i is set, to comprise its valid states dependent on the XInterface (presence of X-IF). i.e. to not consider an RF_WD_COPROC value if there is no X-IF.