What does this pull request do? Explain your changes. (required)

Record and log the following metrics on O:

• Source segment duration
• Download time for source segment
• Transcode score for a segment which is defined as source segment duration / transcode time (this is added for standalone T as well)

B downloads the transcoded results from O separately and the associated HTTP handler lives in LPMS so I did not try to add metrics instrumentation there for now. As a result, this PR does not include the ability to measure the ratio between source segment duration and round trip time from O's POV.

A few other minor updates:

• 1702ba5 uses a NodeType type instead of a string in the monitor package. I felt that this would be less prone to error when writing conditionals based on the current node type as is done in 7a05697 which skips extra state tracking code in SegmentEmerged if the function is used for a node type other than a broadcaster
• 321f5d0 tweaks the view distribution for the source segment duration metric to make the percentile graph of source segment duration easier to understand

Specific updates (required)

See commit history.

How did you test each of these updates (required)

Tested manually and observed metrics using livepeer/docker-livepeer#44.

Does this pull request close any open issues?

Fixes #1661

Checklist:

• README and other documentation updated
• Node runs in OSX and devenv
• All tests in ./test.sh pass