Fix LB transcoder race condition around session-cleanup

[jailuthra]

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

Ensure that no segments are left hanging since we increased the LB channel size in 882ed38

Specific updates (required)

• 415c1c0 - Add a test case to check the case where Seg 1 fails, causing session loop to return, and Seg 2 comes in before the session channels are cleaned up LB transcoder race condition for segment that comes in really fast for the same session after a LPMS transcode error  #1750 (comment)
• 8c6f5b4 - Fix the race condition by adding a separate done channel to signal the sender that the session loop has closed (and return ErrTranscoderStopped to the caller)
• ffd4505 - Fix some old unit tests to use maxSegmentChannels and initialize the done channel

How did you test each of these updates (required)

• Run the test case in 415c1c0 to reproduce hanging behaviour
• Run the test case again after the fix to verify hanging behaviour does not happen anymore

Does this pull request close any open issues?

Fixes #1750

Checklist:

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

[yondonfu]

Looks pretty good. Just a few minor comments relating to the tests.

[kyriediculous]

Looks good to me pending Yondon's comments 👍

[yondonfu]

LGTM after squashing