bump libuv to 1.46.0

[dulmandakh]

there are many bug fixes and improvements, like io_uring support https://www.phoronix.com/news/libuv-io-uring

[achimnol]

Would this enable writing true async file I/O implementation aside the current thread-pool based solutions such as aiofiles? 👀
Since there is no official AbstractEventLoop interface to offer async file I/O yet, I think uvloop may need to define additional methods to expose the underlying libuv's file I/O APIs and write a high-level wrapper as a separate module or package.

[dulmandakh]

Would this enable writing true async file I/O implementation aside the current thread-pool based solutions such as aiofiles? 👀
Since there is no official AbstractEventLoop interface to offer async file I/O yet, I think uvloop may need to define additional methods to expose the underlying libuv's file I/O APIs and write a high-level wrapper as a separate module or package.

I expect that it'll make networking much faster for Python.

[achimnol]

Would this enable writing true async file I/O implementation aside the current thread-pool based solutions such as aiofiles? 👀
Since there is no official AbstractEventLoop interface to offer async file I/O yet, I think uvloop may need to define additional methods to expose the underlying libuv's file I/O APIs and write a high-level wrapper as a separate module or package.

I expect that it'll make networking much faster for Python.

That will be also an expected effect, but what I'm looking forward to see is the native async operations on disk file I/O without using thread pools.

[fantix]

Thank you for the PR! I'm closing this one as #571 is merged. Regarding io_uring-based file I/O, another Python package as a uvloop extension is probably the most feasible way to implement it - I'm open to adding libuv-specific Cython APIs to support extensions.