Bug 1883721 - Move the support_class modify_class call inside the _compose_cmd r=kshampur,perftest-reviewers

Previously modify_class call was right outside of the _compose_cmd call.
Because of that this wasn't shared between the main and extra-profiler runs.
After moving the modify_class call inside of this function, it's now shared
between these two runs since we call this method explicitly for both of these
runs.

Differential Revision: https://phabricator.services.mozilla.com/D207577

Author: canova

testing/raptor/raptor/browsertime/base.py

@@ -599,13 +599,19 @@ def _compose_cmd(self, test, timeout, extra_profiler_run=False):
                 browsertime_options=browsertime_options, test=test
             )
 
-        return (
+        cmd = (
             [self.browsertime_node, self.browsertime_browsertimejs]
             + self.driver_paths
             + [browsertime_script]
             + browsertime_options
         )
 
+        if test.get("support_class", None):
+            LOG.info("Test support class is modifying the command...")
+            test.get("support_class").modify_command(cmd, test)
+
+        return cmd
+
     def _compose_gecko_profiler_cmds(self, test, priority1_options):
         """Modify the command line options for running the gecko profiler
         in the Firefox application"""
@@ -917,10 +923,6 @@ def output_timeout_handler(proc):
         # this will be used for btime --timeouts.pageLoad
         cmd = self._compose_cmd(test, timeout)
 
-        if test.get("support_class", None):
-            LOG.info("Test support class is modifying the command...")
-            test.get("support_class").modify_command(cmd, test)
-
         output_timeout = BROWSERTIME_PAGELOAD_OUTPUT_TIMEOUT
         if test.get("type", "") == "scenario":
             # Change the timeout for scenarios since they