chore: update to lotus 1.5.0-pre1 and actors v3

[iand]

Tests for actors v3 are copy/paste of v2 tests with fixups for the types. The main change for v3 is the structure of the HAMT.

[codecov-io]

Codecov Report

Merging #371 (e5fd69e) into master (0f5a496) will increase coverage by 0.0%.
The diff coverage is 41.1%.

@@          Coverage Diff           @@
##           master    #371   +/-   ##
======================================
  Coverage    43.0%   43.1%           
======================================
  Files          25      25           
  Lines        1917    1932   +15     
======================================
+ Hits          826     833    +7     
- Misses        964     972    +8     
  Partials      127     127           

[placer14]

Confused how the ffi-stub was building before the types or constants were present. I don't see where it was changed but checking out locally it obviously points to that module. Just some small confusion, but looks good otherwise. 🤝

[placer14]

May be nothing, but curious why this was needed?

[iand]

Not my doing. I think gofumpt does it. Numbers starting with 0 are octal in Go, and they added 0o as a consistent way to write them to match 0x

[willscott]

only the subset of ffi stub actually referenced by us or lotus is needed when a module replacement happens.

[frrist]

LGTM! 🙏

[willscott]

ℹ️ There is a dependency on statediff for parsing message parameters. that may also need to get pulled up to the v3 dependencies to continue functioning. I expect to do that this week.

[iand]

Confused how the ffi-stub was building before the types or constants were present. I don't see where it was changed but checking out locally it obviously points to that module. Just some small confusion, but looks good otherwise.

They were in the real ffi package but not imported into Lotus until recently.