Fix reading source file with time attribute

[paulromano]

This PR addresses an issue identified by @yrrepy in #2223. It's a one-line fix, but I've gone one step further to add a test that would have caught it in the first place.

[gridley]

LGTM!

[gonuke]

I don't see how this test actually tests anything? The test reads source.h5 but the PR adds a file with a different name (presumably this new file has a time attribute in the source). Should the test be reading this new source file?

[gonuke]

Is this related to @ypark234 's addition of surface source read/write?

[paulromano]

@gonuke The unit test and the update of the surface_source_true.h5 file are independent of one another:

• The surface_source regression test generates a surface source and the format has now changed since it includes a time attribute, hence the update of the "true" .h5 file for comparison
• The unit test checks that a source file created from the Python API (source.h5) can be used by the underlying C++ code (this test fails without the corresponding fix in the PR)

[gonuke]

LGTM!