Move test timeouts to constants.

[no2chem]

This PR moves timeouts to constants, so that in the future different timeouts can be
set in Travis. It also forces the user to select from a set of four timeout lengths.

[Maithem]

Timeouts are used all over the place, not just with the scheduled executor, but this is a good start. Other timeouts can be discovered and fixed as we go

[slfritchie]

Michael, this patch does something weird with changing the typing of CancellationException. TravisCI (and my run of the suite locally) shows 18 errors, starting with https://travis-ci.org/CorfuDB/CorfuDB/builds/182238645#L1686

[no2chem]

@slfritchie It's not changing the typing of CancellationException. What makes you think that? The patch currently has a bug because it uses getNanos() instead of toNanos(), resulting in a timeout that is 0s, so the tests are cancelled immediately. Fixing soon.

[slfritchie]

Good to know about the units problem.

My question about typing came from this test output:

<<< FAILURE! - in org.corfudb.AbstractCorfuConcurrencyTest
concurrentTestsThrowExceptions(org.corfudb.AbstractCorfuConcurrencyTest)  Time elapsed: 0.404 sec  <<< FAILURE!
java.lang.AssertionError: 
Expecting:
 <java.util.concurrent.CancellationException>
to be an instance of:
 <java.io.IOException>
but was instance of:
 <java.util.concurrent.CancellationException>
	at org.corfudb.AbstractCorfuConcurrencyTest.concurrentTestsThrowExceptions(AbstractCorfuConcurrencyTest.java:37)

[no2chem]

@slfritchie Okay, and that was because the test was expecting an IOException, but the test was cancelled before the exception was thrown...

[dahliamalkhi]

yeh, useful constants. I suspect we'll refine this as we go along, but at this point, any step towards defining and unifying timeouts/iterations as constants is a good step.