Implement HTTP browser caching for blobs

[jbethune]

fix #113

I also checked Starlette's code, but they don't use the Cache-Control header at all.

This PR does not implement caching for language strings because there are no etags for them (yet).
Edit: Basic caching without ETags is now implemented as part of this PR for localization strings.

I tested the code in Firebug and the caching works nicely.

[t-muehlberger]

@jbethune maybe we should not yet close #113 since, as you mentioned, language strings are not yet cached.

Edit: Regarding the missing ETags. As you know you can compute the ETag using any hashing algorithm of your choice, they do not need to be stored in the DB or the YAML-Config. But you should probably do it only once over the lifetime of the backend-process rather than for each request as the hashing can be an expensive computation. You can probably just hold the computed ETags in memory as the number of languages will be relatively limited any you only need a few bytes of memory per ETag. @scy please correct me if I am wrong.

[jbethune]

I can also implement basic caching instructions for language strings without etag is part of this PR. That's actually easy to do. Then each set of language strings can have an expiration of e.g. 1 day.

I would also like to point out that there is #60 but that is in a later milestone. I suggest I implement caching for language strings as described in the previous paragraph and then we implement etag-based caching once we have migrated the translations into a separate location. Then we can calculate a new etag for each version of the translations whenever there is a change.

[scy]

No time for a full review right now, sorry, but about the translation strings caching:

The Config is a singleton and will not change during the application’s lifetime. You could simply use something like

strings = all_frontend_strings(language)
etag = str(id(strings))

because all_frontend_strings employs a lru_cache and should thus respond with the same dict instance per language (please double-check).

If you want to be super careful about id reuse between application runs, feel free to add an application startup timestamp or something to the etag.

[jbethune]

strings = all_frontend_strings(language)
etag = str(id(strings))

because all_frontend_strings employs a lru_cache and should thus respond with the same dict instance per language (please double-check).

Implemented as suggested. This should still be improved when #60 is implemented, because it's just a hack that works most of the time.

[jbethune]

I have rebased this public branch to catch up with other changes and fixed all merge conflicts.

[scy]

Thanks! We might not merge this until after the initial go-live, simply to reduce the number of moving parts, but I appreciate the rebase and will look at it as soon as I can.

[jbethune]

@scy Please let me know when you want to review this, so that I can rebase it again.