If a Schnorr signature aggregation scheme were to be integrated into Bitcoin. Would the reduction in witness data versus non witness data (need less bytes for a signature therefore witness data usage would go down per tx) create a lower upper limit on potential blocksize (believe the realistic upper limit is 2.5 MB rn)?
SegWit blocks aren't limited in bytes anymore but rather in weight. The maximum weight for a block is 4M. The weight of non-witness data is 4x its number of bytes.
So, yes, decreasing the amount of signature-data frees up some weight which can then be filled by more transactions. Yet those new transactions also contain some non-witness data which is more costly. An average full block will still be 4M in weight, but reduced in size (bytes) while containing more transactions. Both are good news.