Working analysis pipeline

[dudewheresmycode]

Getting the core speech-to-text / subtitle analysis pipeline working. This is the process that analyzes a media file and compares with a subtitle to check for accuracy. This is all still pretty rough, but wanted to get a PR going to cap this work at some point.

• Allows selecting movie file and subtitle file for analysis
• Basic (in-memory) job queue system
• Uses ipc to communicate analyze events with the client

I think we'll eventually do this part in the background. I see it working like:

1. User selects/adds a folder to scan
2. We watch for changes, or periodically scan that folder for new media files
3. We analyze any local subtitle files, if none exist we can search for one on OpenSubtitles
4. We verify and "magically fix" subtitles that just need some adjustment

[dudewheresmycode]

Note: This won't currently build correctly without the workaround, or our PR (mozilla/DeepSpeech#3702) being accepted on the DeepSpeech side.

Workaround: I was able to get MacOS electron binaries built on my fork's PR (dudewheresmycode/DeepSpeech#1). Which allows us to get our app building locally on mac.

1. Checkout this branch
2. Download binary for 15.3
electron-v15.3.zip
3. Install it to node_modules/deepspeech/lib/binding/v0.10.0-alpha.3/darwin-x64/electron-v15.3
4. npm start

I added a patch script to our postinstall. So you can just checkout this branch and do a npm install