feat: implement lotus-sim

[Stebalien]

Implements #6156.

Usage

> make build
> cd cmd/lotus-sim
> go build .
> ./lotus-sim create
> ./lotus-sim run

Limitations: to run this, you need 2 weeks worth of state-trees to get power lookback (used to determine per-miner onboarding rates). Ideally, we'd do something else (maybe just look at the number of sectors and make assumptions?

Goals

• Test how the lotus VM and actors as the network grows.
• Find potential bugs that only appear at scale.
• Determine state growth rate.
• Test network upgrades.

Non Goals

This simulation does not:

1. Use real proofs/sectors.
2. Sign anything.
3. Verify any signatures.
4. Use a real miner. It just keeps using the last miner to mine a block (from when the network started).
5. Do any "network" related operations. It just creates and validates (skipping proofs) a fake chain.

Design

This simulation:

1. Forks the network at the current head (or at a desired point, see ./lotus-sim create --help).
2. At every epoch:
   1. Funds the "funding" account (f0100) if necessary (by taking from random actors).
   2. Submits window posts for every "healthy" sector.
   3. Submits (batch) prove commit messages for any outstanding pre-commits that are ready to be proven.
   4. Submits (batch) pre-commit messages in the remaining space.

When necessary, the "funding" account will send funds to actors.

[Stebalien]

To the reviewer: Most of the changes here are just a part of the simulation, but not all:

1. It adds some methods to the actors shims.
2. It exposes the ActorStore from the VM (so we can read intermediate state and, importantly, nonces).
3. It sets ActorErr in one VM.ApplyMessage case where it should have been set but wasn't (not consensus critical).
4. It modifies StateTree.ForEach to iterate over un-flushed actors.
5. It fixes the genesis logic to set the balance (to zero) for some actors. This won't actually affect anything as we default to zero on serialize anyways. Unfortunately, that doesn't help when looking at unseralized actors in ForEach (see 4).

Please pay careful attention to 4. It shouldn't cause any problems (mostly used in genesis and some specific upgrades), and brings the function in-line with expectations. But it's still something that's good to look at carefully.

[magik6k]

Mostly cosmetic nits.

It would also be really great to have at least a basic integration test, so that we know when it would get completely broken by some change.

Should also have a make target, and be added to BINS so it's built in make buildall