feat: multiple editable modules

[brprice]

Support multiple editable modules. Previously one could import as many modules as wanted, but only had one module to write code in. This is now relaxed so writing larger, multi-module programs is possible without abusing imports.

Closes PRIM-23 and PRIM-76

[brprice]

@dhess I'm seeing a unrelated (I think) rel8 test suite failure: https://buildkite.com/hackworthltd/primer/builds/603#67bb9eff-aadb-4da6-8710-a0476d028ce0, which says

    ListSessions
      listSessions:                              FAIL (4.74s)
        Insert all sessions   (2.38s)
        Get all, offset+limit
        Get 25
        Get 76-100
        Get crossing end
          /tmp/tmp-postgres-data-a61157f8685c7193/pg_stat: removeDirectoryRecursive:removeContentsRecursive:removePathRecursive:removeContentsRecursive:removeDirectory: unsatisfied constraints (Directory not empty)
        
        Use -p '/listSessions/' to rerun this test only.

I'll restart the test to see if it is reproducible.
Edit: Retrying the test has passed.

[dhess]

Hmm, that used to occur from time to time due to a race condition in tmp-postgres, but I thought I'd worked around it. Let's keep an eye on it.

[brprice]

This is ready for review. This implements the FR #321

[brprice]

There is something a bit odd in CI: https://buildkite.com/hackworthltd/primer/builds/650#e19733c0-f807-45f2-b4ba-26b67caaf831 "Unexpected lambda in case", and then some GHC Core is output (but CI still passed). I'll investigate, but I don't think this should delay review.

Since CI still passed, I'm not sure whether this is a new symptom. It was just luck that I happened to look at CI to check progress at the right time to notice it. Edit: Looks like it only occured on the last commit (3d54b9f)

[dhess]

I wonder if it's due to a dependency bump via 69cce80

edit: Per discussion offline with @brprice, sounds like it's not, as he hadn't rebased this PR on main when the CI issue occured.

[brprice]

I have rebased onto 69cce80 and it looks like the GHC message has disappeared:

• https://buildkite.com/hackworthltd/primer/builds/665#c977a497-c194-4975-a385-d947b8fbccdd
• https://buildkite.com/hackworthltd/primer/builds/665#9c764e17-c92f-422f-946f-adaffacdf132

I wonder what change caused this (edit: from some investigation, it seems that multiple GHCs will emit this message in certain circumstances, and it is rather sensitive to small code changes.)

[linear]

PRIM-23 Modules phase 2

• Naming modules
• Breaking programs into modules

PRIM-76 Multiple editable modules

[dhess]

For anyone who reviews this after me (including @georgefst), this is one PR where I found it very helpful to turn on GitHub's "Hide whitespace" option, as there are otherwise large chunks of code that are unchanged save for indentation.

[dhess]

I'm afraid I'm too far removed now from Primer core to have anything particularly interesting to say, but this all looks fine to me. Let's obviously wait for @georgefst to weigh in.

[georgefst]

Looks great!

[brprice]

I have rebased onto 69cce80 and it looks like the GHC message has disappeared:

* https://buildkite.com/hackworthltd/primer/builds/665#c977a497-c194-4975-a385-d947b8fbccdd

* https://buildkite.com/hackworthltd/primer/builds/665#9c764e17-c92f-422f-946f-adaffacdf132

I wonder what change caused this (edit: from some investigation, it seems that multiple GHCs will emit this message in certain circumstances, and it is rather sensitive to small code changes.)

I think there is nothing wrong with our code, and am filing it as a ghc issue: https://gitlab.haskell.org/ghc/ghc/-/issues/21555