Full ram for ml-dsa-87 on stm32f4discovery

[dop-amin]

While trying to run benchmarks for ml-dsa-87 I noticed that it doesn't work on the stm32f4discovery.
As we need 129kB of RAM, I think it's required to enable the "fullram" ld script just as for Dilithium5 back then.

[mkannwischer]

Thanks @dop-amin. Sorry I missed that when renaming to ml-das.

I tested that test.py and benchmarks.py works on my DISCO board now. testvectors.py does not seem to work for that implementation.