Generalize file splitting for output writers

[glwagner]

This PR generalizes file splitting for JLD2OutputWriter, replacing max_filesize with a file_splitting keyword argument. To implement a maximum filesize we use

JLD2OutputWriter(model, outputs; file_splitting=FileSizeLimit(max_filesize), ...)

This feature should also be extended to NetCDFOutputWriter in this PR -- hopefully @josuemtzmo can help!

This PR will enable TimeInterval (or any other schedule --- IterationInterval, WallTimeInterval) to be used for splitting output files.

Closes #2967

[navidcy]

We should also add a test similar to the one that was added in #3512.

[glwagner]

We should also add a test similar to the one that was added in #3512.

Doesn't that already exist?

We just have to adapt the test

[glwagner]

Also open to comments about the new names / kwargs.

[navidcy]

We should also add a test similar to the one that was added in #3512.

Doesn't that already exist?

We just have to adapt the test

Yes, the test I think only tests the NetCDFOutputWriter.

[josuemtzmo]

Doesn't that already exist?
We just have to adapt the test

Yes, the test I think only tests the NetCDFOutputWriter.

There is already a test for the JLD2OutputWriter

[josuemtzmo]

I will work on the time splitting schedule. Should we do it in this PR? or should I do another one once it is ready?

[glwagner]

why this change?

[josuemtzmo]

Because path was used later on in the code, but it wasn't updated. Thus the code crashed while creating the file and when the writer.filepath changed but not the path . The easier fix was to replace all instances of path by the writer.filepath.

[glwagner]

Ok, that's a good reason!

[navidcy]

lgtm!

[josuemtzmo]

Why are some of the tests failing?

[navidcy]

Sometimes (often) they just fail to initialize. Don’t assume you broke them necessarily. Uou need to see the log to understand if actually the tests broke or never started.

[navidcy]

@glwagner would you like to have a look at this or should we merge when tests pass?

[navidcy]

Some of the doctests are failing because of the changes:

https://buildkite.com/clima/oceananigans/builds/15164#018e6aa7-67f8-4280-b331-ad7179892dd1/32-18831

[glwagner]

I think it looks pretty good! Thanks @josuemtzmo for the work on NetCDFOutputWriter.

In a future PR, we can add a test to ensure that file_splitting = TimeInterval(T) works.

Also I'm trying to brainstorm a better word to use than "actuated" for talking about scheduling.

But I think once the tests pass we should merge this.