NTT: wrong counting for sampler valid causing deadlock

[mojtaba-bisheh]

If sampler valid de-asserts at data beat number 63 it can cause a deadlock, where the computation never finishes.
Here is the response from the designer:

This is a bug. I see in the waves that, sampler_valid goes low on the very last read. In that case, since data is not valid, NTT should wait for sampler_valid to go high again and then read, execute and write. But in this case, since the read counter is already updated, it thinks that it already saw the last data whereas the write fsm is stuck because it only counted 63 writes instead of 64.

ntt_top_never_leaves_busy3.fsdb.zip