Bug 1864921 - Switch SP3 to use the mean as the summary statistic and do cold iterations. r=sparky,perftest-reviewers

Using the mean matches how upstream speedometer combines the scores.

Running the suite cold is a more realistic configuration because it
avoids using the jithints and any other caches from earlier runs.

Differential Revision: https://phabricator.services.mozilla.com/D193684

testing/perfdocs/generated/raptor.rst

@@ -4255,6 +4255,7 @@ Standard benchmarks are third-party tests (i.e. Speedometer) that we have integr
    * **alert threshold**: 2.0
    * **apps**: firefox, chrome, chromium, safari, custom-car
    * **browsertime args**: --browsertime.speedometer_iterations=5
+   * **cold**: true
    * **custom data**: true
    * **expected**: pass
    * **expose browser profiler**: true
@@ -4266,7 +4267,7 @@ Standard benchmarks are third-party tests (i.e. Speedometer) that we have integr
    * **page timeout**: 600000
    * **repository**: https://github.com/WebKit/Speedometer
    * **repository revision**: 36712deb57d45879f56dd2136f8a6ed8a22d964b
-   * **submetric summary method**: median
+   * **submetric summary method**: mean
    * **subtest lower is better**: true
    * **subtest name filters**: tests/,s3/
    * **subtest unit**: ms
@@ -4597,6 +4598,7 @@ Standard benchmarks are third-party tests (i.e. Speedometer) that we have integr
    * **alert threshold**: 2.0
    * **apps**: fenix, geckoview, refbrow, chrome-m, cstm-car-m
    * **browsertime args**: --browsertime.speedometer_iterations=5
+   * **cold**: true
    * **custom data**: true
    * **expected**: pass
    * **gecko profile entries**: 14000000
@@ -4607,7 +4609,7 @@ Standard benchmarks are third-party tests (i.e. Speedometer) that we have integr
    * **page timeout**: 420000
    * **repository**: https://github.com/WebKit/Speedometer
    * **repository revision**: 36712deb57d45879f56dd2136f8a6ed8a22d964b
-   * **submetric summary method**: median
+   * **submetric summary method**: mean
    * **subtest lower is better**: true
    * **subtest name filters**: tests/,s3/
    * **subtest unit**: ms

testing/raptor/raptor/tests/benchmarks/speedometer-desktop.ini

@@ -26,10 +26,11 @@ unit = score
 browsertime_args = --browsertime.speedometer_iterations=5
 custom_data = true
 owner = Performance Team
+cold = true
 repository = https://github.com/WebKit/Speedometer
 repository_revision = 36712deb57d45879f56dd2136f8a6ed8a22d964b
 host_from_parent = false
-submetric_summary_method = median
+submetric_summary_method = mean
 subtest_name_filters = tests/,s3/
 test_script = speedometer3.js
 test_summary = flatten

testing/raptor/raptor/tests/benchmarks/speedometer-mobile.ini

@@ -25,10 +25,11 @@ unit = score
 browsertime_args = --browsertime.speedometer_iterations=5
 custom_data = true
 owner = Performance Team
+cold = true
 repository = https://github.com/WebKit/Speedometer
 repository_revision = 36712deb57d45879f56dd2136f8a6ed8a22d964b
 host_from_parent = false
-submetric_summary_method = median
+submetric_summary_method = mean
 subtest_name_filters = tests/,s3/
 test_script = speedometer3.js
 test_summary = flatten