drivers: Refactor package as a more generic and performant library

[victorges]

What does this pull request do? Explain your changes. (required)
This is to make some improvements to the drivers package so that it can also be used
as a library from other projects, specifically the task-runner for VOD tasks. It is currently
using the drivers library for a platform-agnostic way of accessing Object Stores for reading
and writing VOD assets.

The greatest advantage of that is that we can just use the same Object Store API from
livepeer-com to declare where assets should be stored as well as stream recordings,
keeping the whole interface leaner.

Specific updates (required)

• Make the Size property optional in our File representation. It is optional in the AWS SDK and is not returned in all implementations of it, like previous versions of Filebase, so it was required at some point (not anymore) but I still think it makes sense to keep it.
• Make the drivers receive an io.Reader instead of the whole buffered file. The reader is compatible with clients that have the whole bytes in memory, but requiring all clients to buffer the entire contents in memory is not ideal. This would have made the VOD tasks much more inefficient, needing to read the entire files in memory before sending to the OS.
• Use an S3 Uploader for uploading files to S3. This avoids the need of buffering the files before uploading, since the simpler upload SDK requires the file to be an io.ReadSeeker due to hashing purposes. The uploader is the suggested solution in the open issue in their SDK repo.

How did you test each of these updates (required)

• Fixed and then ran/passed all tests in go-livepeer
• Used the drivers lib in the task-runner for VOD and already deployed it in production, it's working great
• Did a bunch of performance tests with Storj folks to come up with the optimal configurations for the S3 uploader (which end up being good both for Storj and for GCP)

Does this pull request close any open issues?
Part of livepeer/task-runner#14

Checklist:

• Read the contribution guide
• make runs successfully
• All tests in ./test.sh pass
• [-] README and other documentation updated (not necessary IMO)
• Pending changelog updated

(does it need a changelog update?)

[leszko]

Added 2 comments, other than that LGTM 👍

Also remember to:

• Update CHANGELOG_PENDING.md
• Consider squashing some commits before merging

[codecov]

Codecov Report

Merging #2223 (1f5892b) into master (30b7330) will decrease coverage by 0.06086%.
The diff coverage is 33.33333%.

❗ Current head 1f5892b differs from pull request most recent head 5c28f2f. Consider uploading reports for the commit 5c28f2f to get more accurate results

@@                 Coverage Diff                 @@
##              master       #2223         +/-   ##
===================================================
- Coverage   54.93891%   54.87805%   -0.06086%     
===================================================
  Files             93          93                 
  Lines          19316       19352         +36     
===================================================
+ Hits           10612       10620          +8     
- Misses          8115        8140         +25     
- Partials         589         592          +3     

Impacted Files	Coverage Δ
core/orchestrator.go	78.75723% <0.00000%> (ø)
drivers/gs.go	10.23622% <0.00000%> (-0.16379%)	⬇️
drivers/s3.go	11.08179% <27.27273%> (+0.94093%)	⬆️
drivers/local.go	50.00000% <30.00000%> (-1.11111%)	⬇️
server/segment_rpc.go	78.66667% <50.00000%> (-0.08519%)	⬇️
drivers/drivers.go	27.96610% <66.66667%> (ø)
drivers/overwrite_queue.go	84.48276% <100.00000%> (ø)
drivers/session_mock.go	52.63158% <100.00000%> (ø)
server/broadcast.go	78.24497% <100.00000%> (ø)
server/mediaserver.go	66.13672% <100.00000%> (ø)
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 30b7330...5c28f2f. Read the comment docs.