feat: nv25 and nv26 - fip-0100 (and some FIP-0098)

[rvagg]

Dev builds of builtin-actors and go-state-types. Still needs a bit more work in actors and we still need a migration and invariants in go-state-types and we need a new ref-fvm to go with it. But we can start looking at work in here too.

[rvagg]

Pulled in #12954 to the bottom of this commit stack, but also has a go-state-types that includes filecoin-project/go-state-types#371 and filecoin-project/go-state-types#372 so we'll need a new RC with those.

[rvagg]

Sadly I think I've introduced a flaky that I'll need to make more robust. After 18 runs locally of #12952 I got a fail:

                Error Trace:    /home/rvagg/go/src/github.com/filecoin-project/lotus/itests/daily_fees_test.go:222
                                                        /home/rvagg/go/src/github.com/filecoin-project/lotus/itests/daily_fees_test.go:625
                Error:          Not equal:
                                expected: 9047612
                                actual  : 8583626
                Test:           TestDailyFees
                Messages:       expected miner to have burned 9047612, but it's burned 8583626

We seem to have run the calculation on the wrong side of a deadline so I'll have to figure out how to be more precise.

[rjan90]

Been running some migration benchmarks based on this branch here: #12959 (comment). Both in "offline" and "online" modes.

[Stebalien]

Note: I've confirmed that the v16 bundle matches filecoin-project/builtin-actors@cd43784 (one commit behind the target commit due to filecoin-project/builtin-actors#1656, but it's good enough for an RC).

[rvagg]

I think this is good to merge (rebase merge pls, I think the commit history is fairly clean). I wouldn't mind @Stebalien having a look at my response above to his last outstanding review. We could deal with it, but we're coping with discrepancies between itest environment and mainnet even though it runs in mainnet mode but with 2k sectors (i.e. params_mainnet) and we have a weird genesis with an initial reserve of 700M. So those changes are to adapt to that and bring it close to mainnet.