Fix monolithic push memory issues

[gerrod3]

According to the spec there are two ways to perform monolithic blob pushes:

1. On the initial POST request with the digest query parameter
2. Empty (normal) body POST, then all in the PUT

There's an unofficial third way that podman can apparently do according to our comments which is:

1. Normal empty POST
2. Entire chunk in one PATCH with no Range header
3. Empty PUT

The code should now call our special large chunk handler for all three cases, we were forgetting case number 2. Note that depending on the client they could still cause the server to go OOM if they send ridiculously large chunks in the normal chunked upload path since we read the entire chunk into memory before saving it. There is nothing we can do here as part of the code lives in pulpcore, but honestly what is that client thinking sending such large chunks!

📜 Checklist

• Commits are cleanly separated with meaningful messages (simple features and bug fixes should be squashed to one commit)
• A changelog entry or entries has been added for any significant changes
• Follows the Pulp policy on AI Usage
• (For new features) - User documentation and test coverage has been added

See: Pull Request Walkthrough

[rochacbruno]

@gerrod3 your changes affects the put method, what about the partial_update (patch)?

Also, can we have backport to 2.19

[rochacbruno]

isn't this getting a larger chunk?

[gerrod3]

@rochacbruno Read my initial comment on the different scenarios of chunked upload. If a PATCH chunk upload is sent that is larger than we can handle and doesn't fall under the special podman case then we are out of luck. Someone needs to fix their client to adhere to the spec or set their chunking size to a smaller value.

[gerrod3]

The most we can do is maybe introduce a setting that checks the size of the chunk and if over a certain amount return a 4XX error that the chunk is too large for a PATCH request, use a smaller size.