Fix for `rescript 10.0.1`

[quinn-dougherty]

Taking a quick stab at this: https://github.com/aantron/bisect-starter-rescript/issues/3

when you try rescript 10 with bisect_ppx:

 $ rescript
Dependency on bisect_ppx
rescript: [1/4] src/common/bisect_common-Bisect.cmi
FAILED: src/common/bisect_common-Bisect.cmi
File "/home/.../bisect-starter-rescript/node_modules/bisect_ppx/src/common/bisect_common.mli", line 68, characters 42-53:
Error: Unbound type constructor out_channel

I piggy-backed a flake.nix for a reproducible development environment, mostly for personal use. We can cherry-pick it out of the PR if we don't want it.

[quinn-dougherty]

Confused why out_channel can be unbounded, since it's clearly in stdlib which apparently is open'd at the top of every file. https://v2.ocaml.org/releases/4.13/api/Stdlib.html#TYPEout_channel

@christianoc told me on https://github.com/rescript-lang/rescript-compiler/issues/5596 that some things to do with Pervasives (which is the same thing as Stdlib) were removed from rescript 10.0, I also notice in the rescript-compiler CONTRIBUTING.md file it says that ocaml 4.14 is used, but I don't think that's related.

[quinn-dougherty]

The issue: https://github.com/aantron/bisect_ppx/issues/406

[cristianoc]

Would this same PR work with the older version of ReScript? The new version should only affect compiling ReScript projects. What part is currently failing?

[quinn-dougherty]

Would this same PR work with the older version of ReScript?

I added rescript-9 and rescript-10 as two separate processes to test. I think I expect constraints like "if you're on rescript 9 then your bisect_ppx version has to be <=2.7.1" and "if you're on rescript 10 then your bisect_ppx version has to be >=2.9.1", ultimately.

The new version should only affect compiling ReScript projects. What part is currently failing?

Based on discussion in https://github.com/aantron/bisect_ppx/pull/400, I think CI gradually started failing because the resolver started pulling down later versions of ppxlib. Note that in the last commit to master in march, many jobs were failing.

[dsiu]

Any quick fix for this for ReScript?