feat(miner): externalise SectorOnChainInfo from AMT

[rvagg]

Ref: filecoin-project/builtin-actors#1604

Draft for now, for benchmarking purposes

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 0% with 123 lines in your changes missing coverage. Please review.

Project coverage is 3.39%. Comparing base (5f30872) to head (7bc0344).

Files with missing lines	Patch %	Lines
builtin/v16/migration/miner.go	0.00%	74 Missing ⚠️
builtin/v16/miner/miner_types.go	0.00%	26 Missing ⚠️
builtin/v16/migration/top.go	0.00%	18 Missing ⚠️
builtin/v15/miner/miner_types.go	0.00%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #338      +/-   ##
==========================================
- Coverage    3.39%    3.39%   -0.01%     
==========================================
  Files         662      663       +1     
  Lines      177834   177949     +115     
==========================================
  Hits         6041     6041              
- Misses     169944   170059     +115     
  Partials     1849     1849              

Files with missing lines	Coverage Δ
builtin/v16/miner/miner_state.go	0.00% <ø> (ø)
builtin/v15/miner/miner_types.go	0.00% <0.00%> (ø)
builtin/v16/migration/top.go	0.00% <0.00%> (ø)
builtin/v16/miner/miner_types.go	0.00% <0.00%> (ø)
builtin/v16/migration/miner.go	0.00% <0.00%> (ø)

[rvagg]

Migration takes too long and I run out of memory for this

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
1731013 lotus     20   0  414.0g 169.9g   5.4g S 380.7  90.4  91:32.50 lotus-shed

just before Killed, I don't have enough RAM to support this.

It got to:

2025-01-22T16:46:48.801+1100    INFO    fil-consensus   filcns/upgrades.go:3012 Performing migration: 1184187 of 1185201 jobs processed (749/s) [26m21s elapsed]

Which is 245000 miners processed (I had another log line I didn't push in this commit). So about 1/3rd of the way through in 26 minutes, on a fairly fast machine.

[rvagg]

I think I'd like to explore an approach of updates to the AMT being externalised but existing sectors prior to the network upgrade left as they are. So we'd just have to do some byte sniffing - is it a CID, or is it a SectorOnChainInfo? I believe we've avoided such half-way schemas before, opting to do a full migration into a new schema.