[fs-test-utils]: Use rand_r

Instead of just srand the provided seed, we set it on every SetUp
and in perftests before any test the seed also rests to the given value.
Fixture::seed returns a pointer to the seed that needs to be used by
rand_r.

TESTS: fs-test-utils-test,fs-bench-tests
Change-Id: I59f58d1e901e31211741265d72d8dd3407747c6c

[fs-test-utils] Fix test suite name outputted in the JSON perf results file

Before, the test name was copied as the test suite name, which is not
the right thing to do. An example is
"minfs/Bigfile/16Kbytes/1024-Ops/1-Cycle/Write". When uploaded to the
Catapult dashboard, this gets rejected with a "400 Bad Request" error,
presumably because of the slashes in the name.

This should fix the garnet-x64-perf-swift_canyon bot, which is failing
because its Catapult upload step now fails.

Bug: ZX-2492
Test: "/boot/test/fs/fs-bench-test --out /tmp/results.json && dd if=/tmp/results.json"
and check output
Change-Id: Idd0644ee37e89def020f79f8aa1e6c9fd2f1d490

[fs-test-utils] Remove unnecessary "return;" statements at ends of functions

Bug: none
Test: build
Change-Id: I6e88fa7f44db101251e2984fe101519735923dac

[fs-bench] Return an error if given an unrecognised filesystem type

Bug: none
Test: "/boot/test/fs/fs-bench-test --fs foo ; echo $?"
Change-Id: Ief61aeefc4d5058a8831f5d90a084fc6f34545cf

[fs-test-utils] Fix typo in help text

Bug: none
Test: none
Change-Id: I90c3931389ee277b1af638ec45d18764bea02a5b

[fs-test-utils] Fix help text to match option name

"--runs" is the actual option name, and this matches the perftest
library's "--runs" option.

Note that getopt_long() does actually accept "--run" (because it
accepts abbreviations), though it's not guaranteed to do so if other
options are added with similar names.

Bug: none
Test: none
Change-Id: Ie6e2cee8df11e5829dab3fe443e0e9dd39197ce7

[fs-test-utils]: Add --seed command lined.

Allows easy reproduction of tests that use random number generators.
Each test is seeded individually.

TEST=fs-test-utils-test,fs-bench-test

Change-Id: I2ad80977c42bf78b16f883f77f1ea99f30dafbb8

[fs-test-utils]: Make result_path(--out) optional.

Made the --out flag optiona, but a performance test requires output,
so either --out, --summary_path or --print_statistics MUST be set.
Added test to verify IsValid behaviour.

TEST=fs-test-utils-test

Change-Id: I44be26edcf3595810b22f621d5fca99a2506f358

[fs-test-utils]: Parsing block_device and ramdisk override each other.

Fix error where setting block device via command line flags, due
to default settings. If ramdisk was not explicitly set and
block device is, then clear use_ramdisk.

TEST=fs-bench

Change-Id: I2e64934329b7bd7e082e9aa25e7b1d4f6918c162

[fs-bench] Rewrite fs-bench to use fs-test-utils.

Updated fs-bench to use fs-test-utils, and updated name and
constructs used to be Cpp where possible. Behaviour is
identical, except we sample each read/write op and we can
get estimates from them (average * samples = previous sampled value).

TEST=fs-bench-test

Change-Id: I70a6a1880633355cee65064fa8407831afc10d6e

[fs-test-utils]: `sample_count` fixed per test case.

Test within the same test_case are enforced to have the same
sample_count.

TEST=fs-test-utils-test

Change-Id: I88879b22f463963a702d1656c5cd52d07d3402e5

[fs-test-utils] Add perftest lib.

Provide utility classes to run a perftest test, allowing the underlying
benchmark to be controlled via customized options.

Test methods are given access to Fixture, which can be used to obtain
information about the block device path, etc and the lib
will call SetUp/TearDown(TestCase) accordingly also allowing
test to run in sequence without tearing down between each test,
in case Test depends on each other. (E.g. Test write, and then read
on the written stuff).

TEST=fs-test-utls-test

Change-Id: I5127d20e55f4a24d81b3a56a6c30b6e802a69948