ice40: simulate SPRAM poweron delay in cells_sim

[sapphire-arches]

Makes the SB_SPRAM simulation model slightly more accurate, including a coarse simulation of how the block behaves when modifying the various power/sleep state control lines.

I'd also like to write up some tests for this, but the actual modifications to the cell simulation should be ready for review.

[smunaut]

Mmm, I don't think that logic is correct. If you release power poweroff and then release sleep slightly after, you wouldn't see the 41 ns delay, because state would go POWEROFF -> SLEEP -> NORMAL.

[sapphire-arches]

I've reworked this to use a handful of counters instead of a state machine, which was getting extremely unwieldy when trying to account for the various ways the power control lines can interact.