Add watched directory functionality

[ianalexander]

This PR is a followup to issue #465.

This PR will add watched directory functionality to OCRmyPDF. This PR includes the following changes:

1. watcher.py: This file uses python's watchdog to watch a folder for new files, and then immediately send them to OCRmyPDF. Results are stored in a separate output folder. Optionally, output folder files may be organized by year and month. These are all configurable by environment variables.
2. Dockerfile: We now include the new requirements (listed in requirements/watcher.txt) in building the docker image. This docker image can now be launched directly with the watcher.
3. docs: I've updated the existing watched folders section to now reference the docker image with the watcher.py script. This includes examples of usage.

Here is how you test this:

git clone
cd OCRmyPDF
docker build -f .docker/Dockerfile .
docker run \
   -v $(pwd)/tmp/incoming/:/input \
   -v $(pwd)/tmp/:/output \
   -e OCR_OUTPUT_DIRECTORY_YEAR_MONTH=1 \
   -it --entrypoint python3 \
   <container id> \
   watcher.py

After this, drop a pdf into ./tmp/incoming/. Output should look something like:

Starting OCRmyPDF watcher with config:
Input Directory: /input
Output Directory: /output
Output Directory Year & Month: True
New file: /input/input.pdf.
Attempting to OCRmyPDF to: /output/2019/12/input.pdf
Scan: <snipped>
OCR: <snipped>
...

Afterwards, verify in ./tmp/ that the output file was OCR'ed correctly.

What do you think?

[ianalexander]

Good feedback. I've updated the code to use pathlib instead of os and included the old content in the docs.