[Feature Request] Video frame extraction

**Is your feature request related to a problem? Please describe.**
Video data is often a useful source of images, but can be difficult to get synchronised with corresponding telemetry and other data. It is also frequently challenging to efficiently get "good" frames, by frame quality, by physical spacing, etc

**Describe the solution you'd like**
- [ ] extract timestamped frames from [QGroundControl-saved video](https://www.youtube.com/watch?v=gUxlby1Dznk) (and other formats?)
- [ ] a method for aligning video frames with corresponding telemetry and other data
- [ ] a method of extracting the best (sharpest?) frame within a small time or area region
- [ ] intelligent frame extraction that avoids large vibrations/acceleration
- [ ] a method of extracting frames by distance travelled since last frame
- [ ] a method of extracting frames in specific regions (waypoints)
- [ ] a way of stitching frames together into a composite image
- [ ] a way of aligning frames or composite images onto/into a map

**Describe alternatives you've considered**
| alternative | reason(s) why unsuitable |
|---|---|
| manual frame extraction | slow, time-consuming, impractical |
| direct time-based extraction (e.g. ffmpeg, every nth second) | imprecise, may get a bad frame even if the next one was great, doesn't make intelligent use of other available data |

**Additional context**
- [`VideoReader`](https://github.com/ES-Alexander/pythonic-cv#videoreader) - nice setup, efficient processing
- [initial stitching work](https://github.com/ES-Alexander/pythonic-cv/blob/add-stitching-example/src/pcv/examples/stitching_advanced.py) (upstream, in `pythonic-cv`)