change `namespace` annotation to `is_namespace` and use `namespace` for name of namespace
dont change commit when commit is None - otherwise always reset to recent version when we don't intend to
Use a MetaPathFinder to allow python to resolve imports from the temporary model directory
Build dependent versions in split builds
Clear package imports when allow_repo=False
Just going to make a special completion phase for the top-level NWBFile because it's so special cased it's not worth trying to generalize.
- Resolving hardlinked datasets in read maps
- Resolving 2nd-order empty groups
- Completing container groups
- Move fixing references to recursive function
- Multiple passes for a given phase
- NDArrayProxies should resolve dask or h5py reading on their own
- Correctly build split schema from PydanticProvider instead of needing external logic
Track dependencies in build results and dont try and complete model if dependencies not alreay completed (ie. need multiple passes to complete, which is the correct behavior)
Use base models in dynamic tables, we should fix the dynamictable base class rather than just having everything be an anonymous BaseModel in order to satisfy type validation
Clear completed queue when advancing stages - we should make sure to move everything to completed with maps rather than implicitly keeping things in completed when they aren't - so we might be getting old versions of results that aren't actually fully completed but have some reference stubs still in them.
