Remove spacetimedb-core as a dep of cli

[coolreader18]

Description of Changes

This is possible now that standalone is now a separate binary from cli. This lets us separate concerns, and speed up compilation of the cli.

Expected complexity level and risk

1 - moves some stuff around but otherwise a pretty small change.

Testing

• No functionality change, tests still pass for codegen for our test modules (for rust, csharp modules).

[mamcx]

Why was removed the TokenValidationError?

[coolreader18]

This method never actually returned anything other than its Other(anyhow::Error) variant, so it's easier to just return anyhow::Error.

[bfops]

why is the feature newly needed here?

[coolreader18]

It was being enabled by core, but now core isn't in the build graph.

[bfops]

why is this feature needed now?

[coolreader18]

Same reason - no longer being enabled by some other crate now that core isn't in the build graph.

[bfops]

I'm broadly in favor of improving compile times. +1 to Mario's questions.

Dumb question - what exactly is spacetimedb-core? At first glance, it seems entirely reasonable for the CLI to depend on a core spacetimedb library.

I don't personally feel equipped to confirm that the changes don't do something meaningful, particularly around wasmtime. Can you add a bit more detail to the Testing section about the existing test coverage for generate?

[coolreader18]

spacetimedb-core is the core of the database implementation, and now that is being pulled in by standalone. The CLI only needs to be able to talk to a database, but it doesn't need to run one itself.

[coolreader18]

I could copy over Mem and MemView from spacetimedb_core::host::wasmtime, but that would mean there's more code and unsafe duplicated overall, and this only uses a small amount of its functionality.

[coolreader18]

The internal tests are only because Cargo.lock needs regen. I feel like those tests probably shouldn't use --locked