Leverage covergroups to generate data

[insoochung]

• [x] Covergroup parser
• [ ] Vectorize covergroups

[insoochung]

@abucturab, lemme know when you get the covergroup parser done.

And as for now, I think feeding CDFG of covergroup definitions as an extension of the design CDFGs should be ok. I've done some research and some successful usages take concatenated image patches and text embeddings as input (keyword: multi-modal transformer). One concern is that we use way less data. However, we do incorporate pre-trained model as text readers so I hope it compensates somehow...