Increase rtl and fifo-depth sim timeouts (fixes #1457)

[LeahLS]

Slightly increased the rtl and fifo-depth simulation timeouts to fix issue #1457 to not run into simulation timeouts when building the upscaled bnn-pynq finn-examples for the AUP-ZU3 board.

[fpjentzsch]

FYI: We had similar problems but worked around them differently, by letting LIVENESS_THRESHOLD (defaults to 1M) override this heuristic if it is larger: eki-project#115

[auphelia]

Thanks for pointing this out @fpjentzsch. We’ll proceed with @LeahLS's version for now because we want to avoid overriding behavior that we actually want to observe. If we replace the estimation with 1M, we might miss cases where a component changes the cycle count significantly and our estimation becomes too inaccurate.

[fpjentzsch]

@auphelia Okay, although I think we should ideally detect mismatches between actual and estimated latency/interval via some kind of is-close comparison in the PyTests or in separate regression tests, not via detecting a simulation timeout.