Make calling io->open() explicit

[oruebel]

While working on debugging #123 I came across the issue that open seemed to be called more than once on the same file. Part of the issue seems to be that io->open is being called implicitly on behalf of the user in NWBFile.initialize. This PR updates the logic to always require that the user call io-open before making any modifications to the file, e.g, initializing the NWBFile. This PR makes a few changes:

• Modified NWBFile.initialize to more robustly check if a file has been initialized. The current logic requires that a file must not exist on disk before calling NWBFile.initialize, i.e., if a user calls io->open() before calling NWBFile.initialize, the current logic would fail.
• Added NWBFile.isInitialized to check whether the expected file structure exists within the file
• Updated HDF5IO.create* methods to perform the more strict check that canModifyObjects() is true rather than just whether the file isOpen()
• Updated test suite to call io->open() directly where it was missing
• Updated testWorkflowExamples.cpp to add REQUIRE checks . The example previously passed independent of whether the workflow was successful or not because the status of I/O operations was not being checked. Since the AqNWB functions did not raise exceptions (but simply returned a Status::Failure) the workflow would run to completion even if the file was not being created at all.

[oruebel]

@stephprince this PR is ready for review

[stephprince]

These changes look good to me and I'm fine to make io->open() explicit.

However, I'm not sure what the original issue was with calling io->open() twice? I thought it was handled in Status HDF5IO::open(FileMode mode) that if the file was already opened and io->open() was called, nothing would happen.

[oruebel]

However, I'm not sure what the original issue was with calling io->open() twice? I thought it was handled in Status HDF5IO::open(FileMode mode) that if the file was already opened and io->open() was called, nothing would happen.

I'm not 100% sure yet either. I'm still debugging the Windows issue. But making it explicit at least makes debugging easier.