[switch-optimization]: wrsw_nic/nic_elastic_buffer: use native almost_full and almost_empty for generating stall

Xilinx FIFOs have native almost_empty and almost_full outputs driven at
configured thresholds. Using them saves few LUTs comparing to stall
generation based on usedw counter as it was done earlier.

It's just a trivial clean up commit, it does not save any resources.

The values used by WR Switch software fit in 8-bit registers.

[sim/testbench] adding possibility of easily settin many untagged VLANs to EP driver + some usecases to main.vs

[swcore/rtu] added abort from input_block@swcore to rtu - in case where frame is dropped (ouput stuck) and rtu is processing when we finish with the frame... we abort the decision

[bare_top] many changes which should be in different commits: added MUX to chipscope, added out_blk_states to HWDU, changed MPM parameters - was too slow

[swcore/in_blk] bugfix: when rtu drop decision and EOF happened in the same cycle... we were in trouble

[rtu] abort of rtu_req from EP added - in case of error in the input data, error status is sent to SWcore, abort needs to be sent to RTU, otherwise SWcore gets lost

[swcore/out_blk] dropping frames no retry request from EP + changed IFG cause it was too high (by 1 cyc) when tagging/untatting

[swcore/mpm] problems wtih keeping timing by RR arbiter->replaces by crude arbitration: shiftbuffer->merry goes around -> wroks