Add Browser support for FPL

[jafeltra]

Description: This PR adds support for using FPL in a browser and using the browser's IndexedDB as a PackageCache.

The BrowserBasedPackageCache class is a new PackageCache implementation that can be used to store resources in the browser's IndexedDB. Because the IndexedDB API is asynchronous, the BrowserBasedPackageCache does some extra work to ensure the synchronous methods of the PackageCache interface still work. There is an async initialize function that looks in the IndexedDB to check what packages are already loaded and saves that information to a local map, and the cachePackageTarball function stores resources in IndexedDB and to the local map. The local map is then used for the rest of the methods, like isPackageInCache, as those are all synchronous.

There were also a few other changes that were needed when using in a browser, some of which I'd like your opinion on if there's a better way around it:

• https.get vs axios.get: I could not figure out how to have axios.get return something that could be unzipped in the browser. I suspect I must have missed something, but I ended up using https.get, which is what FSH Online currently uses. If anyone reviewing this has an idea of how to configure axios to get it to work with the tar-stream library or anything else to successfully unzip and get the data without a filesystem, let me know.
• useHttps option in FHIRRegistryClient: When I needed to use https.get instead of axios.get, I had originally created a separate RegistryClient (a BrowserFHIRRegistryClient), but all the code was the same except for the actual GET request. Instead, I ended up adding an option to the existing FHIRRegistryClient in an effort to duplicate as little code as possible. I'm not sure if this is the best approach, so let me know what you think.
• nock for the https.get calls: I could not figure out how to use Jest and spyOn and mocked implementations to successfully return responses that were like what https.get returns. I ended up using the same approach that FSH Online used, which used nock. That does add another dependency, and it feels like Jest should be able to do the same, so if you know of another way around using nock, let me know.
• initialize(initSql) in SQLJSPackageDB: It turns out that I needed to pass a configuration option into initSqlJs() when initializing in the browser, so I needed to be able to initialize it in FSH Online and pass it in to SQLJSPackageDB.

Testing Instructions:
Make sure the tests reflect the expected functionality. I also pushed up a new-fpl branch on FSH Online that uses this branch. Feel free to try out using FSH Online and making sure you can download and load packages.

Related Issue:
Fixes #29

[cmoesel]

This is looking good. Given how similar the registry code is, I think I agree with your general approach not to create another class.

I've left a few comments, questions, and suggestions for your consideration.

[jafeltra]

@cmoesel - I think 73ee511 addresses your comments so far. I answered one of your questions above, but happy to chat more about it.

Two other questions I have that came up while I was working on that commit:

1. I tried to actually spy on the initSqlJs to make sure the options were constructed correctly in the SQLJSPackageDB.initialized test, but I couldn't figure out how to correctly spy on it in the test. If you have any ideas about how to do that and think it is a good thing to test, let me know.
2. When I was running the tests again, I was thinking about the skipped test that I added in the BrowserBasedPackageCache suite. Is it worth having 1 skipped test in our skip report (since this is the first) just to highlight the differences between the DiskBasedPackageCache and that one? Or should I just get rid of it?

[cmoesel]

Thanks for the updates! Regarding your last two questions:

1. I just spent some time trying to figure out how to spy on the initSqlJs function and I also couldn't figure it out without doing some funky things in the production (and test code). I don't think it's worth jumping through hoops.
2. I don't have a strong opinion on this, but I understand your point -- and I think I agree. We probably don't need that skipped test at all, so feel free to delete it.

[jafeltra]

2. I don't have a strong opinion on this, but I understand your point -- and I think I agree. We probably don't need that skipped test at all, so feel free to delete it.

I commented the test out with a comment in f64fc81.

[jafeltra]

I found one other issue when using this in FSH Online that I fixed in 2617104. That commit rejects any packages that encountered an error or an error status code (like if a package can't be found because you made a typo in the Configuration in FSH Online, just as a random example and not because I just did that...). This way the error is correctly caught and handled and other packages can be downloaded.

[mint-thompson]

One minor typo, other than that, everything is in good order!

[cmoesel]

Looks good! Thank you, Julia. It's really neat to have this built into FPL now so that others can use it!

[mint-thompson]

hooray!!!