ci tests are failing for Nvim nightly

[justinmk]

example: https://github.com/neovim/go-client/actions/runs/9061896916/job/24894617688

partial list of failures:

2024-05-13T11:14:00.4084360Z     --- FAIL: TestAPI/Command (0.00s)
2024-05-13T11:14:00.4084900Z         --- FAIL: TestAPI/Command/Commands (0.00s)
2024-05-13T11:14:00.4085893Z             --- FAIL: TestAPI/Command/Commands/Nvim (0.00s)
2024-05-13T11:14:00.4086628Z             --- FAIL: TestAPI/Command/Commands/Batch (0.00s)
...
2024-05-13T11:14:00.4106998Z     --- FAIL: TestAPI/Key (0.01s)
2024-05-13T11:14:00.4107493Z         --- FAIL: TestAPI/Key/Nvim (0.01s)
2024-05-13T11:14:00.4110940Z             --- FAIL: TestAPI/Key/Nvim/KeyMap (0.00s)
2024-05-13T11:14:00.4112397Z         --- FAIL: TestAPI/Key/Batch (0.01s)
2024-05-13T11:14:00.4115897Z             --- FAIL: TestAPI/Key/Batch/KeyMap (0.00s)

[garyburd]

TestAPI/Command/Commands/Nvim: Test fails decoding a msgpack nil to the Command.Complete field. Is nil and expected value for "command"?

TestAPI/Highlight/Nvim: The test expects the ctermfg and ctermbg fields to be set in nvim_get_hl result, but the fields are not set. Needs more investigation.

TestAPI/Extmarks/Batch: It looks like the test makes assumptions about how nvim assigns ext mark ids. If so, the test is fragile and should be rewritten.

TestAPI/Key/Nvim/KeyMap: nvim_get_keymap returns more results than expected by the test. The test expects different results based on the nvim version. The test is clearly fragile and should be rewritten.

The tests should confirm that the package correctly encodes API arguments and decodes results, but no more.

[zchee]

First, sorry for no sense testcase. But yes, cause by nvim version difference. Therefore currently nightly jobs not requires for merge pull request.

I'll fix when I have time for support to current nightly nvim APIs with respect you said:

The tests should confirm that the package correctly encodes API arguments and decodes results, but no more.