Make parquet support optional for datafusion-common crate

[jonmmease]

Rationale for this change

Hi! This is a small PR that makes the parquet dependency of the new datafusion-common crate optional. The motivation is that I'd like to depend on datafusion-common in the WebAssembly logic of VegaFusion, and the parquet dependency is the only thing preventing datafusion-common from compiling to WASM.

What changes are included in this PR?

The parquet dependency in datafusion-common has been marked as optional, and feature gates are used in the error.rs file (following the pattern of the optional avro dependency). This feature is then enabled by the datafusion crate.

Are there any user-facing changes?

No

[jonmmease]

Thanks for taking a look @Dandandan! All green now

[Dandandan]

This is I think a uncontroversial change, so going ahead and merge this.
Thanks @jonmmease for the contribution. VegaFusion looks really interesting!

[alamb]

This looks good. Very cool @jonmmease

FWIW it might make sense to add a test for this to CI (e.g. something like cargo check --no-default-features -p datafusion-common so that we don't accidentally introduce a dependency again)

[jonmmease]

Thanks @alamb. That's a good idea. I'll play with it.

The other thing to test would be to check that the crate compiles successfully to the wasm32-unknown-unknown target with wasm-pack. Would that be reasonable? Over time, I would be interested in helping to get more of DataFusion working in the browser (c.f. #177), and having the smaller crates now seems like a good opportunity to make progress toward that goal incrementally.

[alamb]

The other thing to test would be to check that the crate compiles successfully to the wasm32-unknown-unknown target with wasm-pack. Would that be reasonable?

Yes definitely! Thank you