Restore tracing in Snitch DMA

[colluca]

Contributions

• Restore DMA tracing in Snitch, to the extent this feature had before the old DMA was replaced in hw: Update iDMA and remove future source directory snitch_cluster#108. Note this does not restore all previously logged signals as significant changes were made to the backend.
• Correct a potentially dangerous typo in the tracer RTL (SYNTHESIS being misspelled).
• Enable tracing multiple DMA channels, when present.
• Add backend signals required to parse the DMA traces in Snitch, as used in trace: Add functionality to parse DMA trace snitch_cluster#71. These are equivalent substitutes to some of the previously logged signals in the Snitch DMA.
• Tentative support for tracing in Verilator (see below).

Tracing in Verilator

Before the iDMA integration in Snitch I was able to enable DMA tracing also in Verilator (version 4.110). With the iDMA this fails with the following error:

%Error-UNSUPPORTED: /scratch/colluca/workspace/snitch_cluster/traces-dma/working_dir/idma/src/frontend/inst64/idma_inst64_top.sv:530:60: Unsupported: Non-single-bit pos/negedge clock statement under some complicated block
  530 |             @(posedge gen_backend[c].i_idma_backend_rw_axi.clk_i); 
      |                                                            ^~~~~

Upgrading to a newer version of Verilator (tested on 5.006) would solve this, but the --timing flag needs to be provided (given the @ timing block, whereas in the legacy DMA the tracer used an always_ff block).

This still fails on the zero-delay statements:

%Error-ZERODLY: /scratch/colluca/workspace/snitch_cluster/traces-dma/working_dir/idma/src/frontend/inst64/idma_inst64_top.sv:530:9: Unsupported: #0 delays do not schedule process resumption in the Inactive region
  530 |         #0; 
      |         ^

but if we guard these with `ifndef VERILATOR, thereby omitting the zero-delays altogether for Verilator, all works as expected.

The question is if we can accept to support only Verilator >5.0, or if we would need to implement version-based guarding. AFAIK, Verilator does not define a VERILATOR_5 macro itself, so we would need to implement something like that ourselves.