This is needed by a customer with a nontraditional codebase.

Add some pollset benchmarks

Fix flush read

Upmerge v1.2.x

Use the right encoding overhead number in remote interop test with Cronet

Don't track allocations if we are not going to use the tracking results

Track memory as part of microbenchmark profiling runs

Skip split_definitions_test if PyPy is used as an interpreter.

PHP benchmarking performance client, proxied through Ruby

Temporarily disable resource quota QPS tests

run_tests_matrix: ignore the number of failed skipped tests

Fix sanity

jobset.run was changed on 16 October 2015 in e4c35610
to return the number of failures and the result set. On 20 October 2016
in 5c691c63 jobset.run was changed to
allow jobs to be skipped, but the introduced code path only returned
the result set, giving jobset.run inconsistent return value semantics.
On 24 October 2016 7e9bd6ca introduced
a use of jobset.run's "skip_jobs" return value semantics, and on 8
March 2017 d159af31 resolved
jobset.run's inconsistent return value semantics but failed to correct
run_tests_matrix's use of the inconsistent "skip_jobs" return value
semantics.

Skip SameSeparateTest, SameCommonTest, SplitCommonTest, and
SplitSeparateTest if the interpreter is PyPy.

Document status ordering rules

Sanitize before bigquery upload

Enhance bm_diff

Bumped version to 1.3.0-dev

Handle floats in bm_json

Change execution order of microbenchmarks

Fix chttp2 ping parser

Finish streams that error. Don't histogram failed ops.

Give execute permission for Kokoro shell scripts