Celux

- Some Major Refactoring and changes made.
- Parsed and created `Filter` classes for every (video) Ffmpeg filter.
- Filters defined within `Celux.pyi`
- Not all are tested. For Full documentation of arguments and usage, see: [ffmpeg-filter-docs](https://ffmpeg.org/ffmpeg-filters.html)
- Please create a new issue if any problems occur!
- Fixed an issue with Filter initialization and unwanted output messages.
py
from celux import VideoReader, Scale , CreateFilter, FilterType


scale_filter = Scale(width = "1920", height = "1080")

scale_filter = CreateFilter(FilterType.Scale)
scale_filter.setWidth("1920")
scale_filter.setHeight"1080")
scale_filter.setFlags("bicubic")

with VideoReader("input.mp4", device = "cpu", filters = [scale_filter]) as reader:
for frame in reader:
will be a scaled frame

- Added some safety checks for filters.
- Fixed issue that occurs when using `scale`.

- Implemented filters for `cpu` usage.
- usage should be familiar to those who've used `ffmpeg`:
py
filters = [("scale", "1280:720"), ("hue", "0.5")]
reader = cx.VideoReader("/path/to/input", device = "cpu", filters = filters)

- Removed Stream Parameter in `VideoReader`: The `VideoReader` no longer accepts an external CUDA stream.
- Introduced event-based synchronization between frame reading operations to ensure proper and consistent output.
- Use of `nvdec` directly.

- Testing some changes, partial release, may end up reverting.
- Use `nvdec` directly instead of `_cuvid`.
- Some small refactoring and testing, nothing major.

- **New Features**:
- Added `num_threads` arg to control decoder threads internally used.
- Fixed `VideoReader()` calls, now properly sets frame range.
- *Potentially* fixed issue with cuda synchronizations.