Reduce reliance on randomness in tests

[johnbillion]

39 of core's tests rely on rand_str() returning a different value each time it's called in order to pass. It's entirely possible that rand_str() could return an identical value twice, and therefore the test would randomly (ho ho) fail.

Let's remove all the unnecessary uses of rand_str(), where a fixed string would suffice. This will also shave off some milliseconds.

[johnbillion]

37371.patch​ replaces all instances of rand_str() with static strings where the test will fail if rand_str() were to return sequential identical strings.

[johnbillion]

In 38382:

Build/Test Tools: Remove many unnecessary calls to rand_str() which can, in theory, fail at random. Static strings are much more appropriate.

See #37371

[jorbin]

@johnbillion Do you want to remove the rest of the calls to rand_strl? If so, good first bug?

[johnbillion]

There's also a bunch of instances of rand() in tests which are just as unnecessary.

[johnbillion]

In 38762:

Build/Test Tools: Begin eliminating unnecessary randomness in tests.

Although unlikely, clashes in randomly generated strings could cause unexpected failures. In addition, most randomness is entirely unnecessary, is bad practice, and increases test time (however small it may be).

See #37371

[johnbillion]

In 38763:

Build/Test Tools: Continue eliminating randomness in tests.

See [38762]
See #37371

[johnbillion]

In 38938:

Build/Test Tools: Continue eliminating randomness in tests.

See #37371

[johnbillion]

In 39556:

Build/Test Tools: Remove some more randomness.

See #37371

[desrosj]

@johnbillion is there any more work to be done here? Can we wrap this up during 5.4?

[davidbaumwald]

@johnbillion Is this still on your plate for 5.4?