Better testing for NumPy #135
Description
As @ryanking13 mentioned in #86, it would be nice if we could test NumPy's full test suite with the Pyodide runtime, or at least a significant portion of it, as an additional source for catching any bugs outside of NumPy's out-of-tree Pyodide CI job that tests only stable Pyodide versions available via cibuildwheel.
I think the best approach would be to create a separate job that can be scheduled nightly and generates issues with https://github.com/xarray-contrib/issue-from-pytest-log if any of the tests fail. It can test against the pinned nightly Pyodide runtime, and the latest nightly Pyodide runtime available in two jobs. We should also make it possible to trigger the job on demand (workflow_dispatch:) and also on encountering a [numpy] commit message, a test numpy label, or a [numpy] in the PR title.