Cleaner nextpnr BBA generation

[gatecat]

Currently the nextpnr BBA generation code includes a lot of duplication, particularly where lists are concerned. e.g. patterns like https://github.com/SymbiFlow/python-fpga-interchange/blob/6a80233ca97427c9b1a1cdd16349c3724914d008/fpga_interchange/chip_info.py#L658-L672 crop up a lot

I think a custom decorator might be the nicest solution here, although I don't have enough Python experience personally to comment on how much work it would be to implement this.

Longer term, the structures in nextpnr (https://github.com/YosysHQ/nextpnr/blob/master/fpga_interchange/chipdb.h) and Python should come from the same source to reduce the risk of mismatches, although I don't know which should be definitive.

[kowalewskijan]

As for me, I think an auto-alignment feature is worth to add too (as a part of this issue) to avoid manual padding for some fields like it is done in BelInfo.

[kowalewskijan]

Longer term, the structures in nextpnr (https://github.com/YosysHQ/nextpnr/blob/master/fpga_interchange/chipdb.h) and Python should come from the same source to reduce the risk of mismatches, although I don't know which should be definitive.

I think the header (chipdb.h) could be generated as a part of nextpnr_emit.py script. Currently in the Nextpnr we call the nextpnr_emit.py from the cmake during bba generation - probably this is the best place to generate the header as we won't re-generate the bba file too frequently.