Handle slicer Toxic with zero SizeVariation

[areveny]

This change skips randomized size variation for the slicer toxic if the SizeVariation parameter is 0, which is also the default setting.

Additionally, it fixes a small bug in the randomization. rand.Intn(1) was being added to mid to adjust for integer division rounding. However, rand.Intn(n) returns an integer in the range [0, n), so rand.Intn(1) always returns 0. It has been changed to rand.Intn(2).

Closes #178

[areveny]

Removed the use of labels in the tests, also reduced the waiting time for timeout so the tests wait less long to complete. (But much more time than the delay so it won't time out unexpectedly.)

[miry]

@areveny It looks like I still have problems random failing tests for Slicer.

$ MallocNanoZone=0 go test -v -race -timeout 1m ./...
...
=== RUN   TestSlicerToxic
    slicer_test.go:53: Expected to read about 480 times, but read 0 times.
    slicer_test.go:56: Server did not read correct buffer from client!
--- FAIL: TestSlicerToxic (0.02s)

[areveny]

Reverted the timeout to its original duration.

[miry]

Thank you @areveny. Can you please update the title of the current PR and entry in CHANGELOG to make it more informative. Example: Handle toxic slicer when SizeVariation is null.
It would remove duplication in history for readers like: Fix bug.

[areveny]

Great. All feedback so far has been incorporated.

[Strech]

LGTM 👍🏼

I would suggest to take a look at the comment which might be outdated

[miry]

@areveny Propose to change the code a bit. Moved the +{0,1} to default, according the comment it was connected to part (end-start)/2. Also removed the comment.

[miry]

Some small refactorings

[areveny]

I have gone with the approach of removing the rand.Intn(2). The comment is removed as well.

[Strech]

Looks cleaner 👍🏼

[neufeldtech]

Changes have been release as part of https://github.com/Shopify/toxiproxy/releases/tag/v2.4.0