[Feature] fully unroll NN models

[walkieq]

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

For some low latency applications, we hope the final NN can be fully unrolled.
Is it possible that Allo could generate such HLS-based designs?

Describe the solution you'd like
A clear and concise description of what you want to happen.

I guess a simple solution is to just place a pragma of "HLS pipeline" in the top module. But I would like to seek your advice.

Additional context
Add any other context or screenshots about the feature request here.