First draft of release notes for v3.3

[cary-ilm]

No description provided.

[lgritz]

Maybe we can summarize why this is important? Is there a perf benefit?

[cary-ilm]

Kimball said he's got notes, I left that section for him

[kdt3rd]

Some notes, not exhaustive:

• The C++ API has started to use the OpenEXRCore library underneath
• This is a transparent change to the existing API, although the ABI (i.e. structure / class layout) has changed
• Existing reading of pixel data should be more efficient due to fewer memory allocations / frees during the process of reading. Additionally, some more specialisation of unpacking routines may result in faster unpack times
• All compression routines are implemented by the C Core layer underneath and no longer duplicated
• Initial support for "stateless" reading of scanlines has been proposed, allowing multiple threads to read scanlines into different frame buffer objects at the same time. While well tested at the Core level, the C++ api should be considered experimental for this release
• Thread dispatch for reading different file types has been made more homogeneous, so is simpler and more consistent

[meshula]

"manifest" is going to be new for some people, wonder if we should have a note for learning more about it?

[meshula]

something like "... of the embedded manifest. The manifest provides a mapping to a human-readable identifiers, see https://github.com/AcademySoftwareFoundation/openexr/blob/25c2d775f889948d48114f382b58e35d313ecad3/website/DeepIDsSpecification.rst for more information."