Create the iterations file after benchmark_run_dir is created.

Change pbench-uperf and pbench-user-benchmark.  This makes all the
current unit tests pass.

agent/bench-scripts/pbench-uperf

@@ -445,7 +445,6 @@ else
 fi
 # we'll record the iterations in this file
 benchmark_iterations="${benchmark_run_dir}/.iterations"
-> ${benchmark_iterations}
 mdlog=${benchmark_run_dir}/metadata.log
 
 function record_iteration {
@@ -466,6 +465,10 @@ function record_iteration {
 }
 
 mkdir -p $benchmark_run_dir/.running
+
+# now that the benchmark_run_dir directory exists, we can initialize the iterations file
+> ${benchmark_iterations}
+
 # save a copy of the command, in case the test needs to be reproduced or post-processed again
 echo "$script_name $pbench_cmd" >$benchmark_run_dir/$script_name.cmd
 chmod +x $benchmark_run_dir/$script_name.cmd

agent/bench-scripts/pbench-user-benchmark

@@ -128,7 +128,7 @@ benchmark_run_dir="$pbench_run/${benchmark_fullname}"
 
 # we'll record the iterations in this file
 benchmark_iterations="${benchmark_run_dir}/.iterations"
-> ${benchmark_iterations}
+
 mdlog=${benchmark_run_dir}/metadata.log
 
 # make the run dir available to the user script
@@ -137,6 +137,9 @@ benchmark_results_dir="$benchmark_run_dir/$iteration/sample1"
 mkdir -p $benchmark_results_dir/.running
 export benchmark_results_dir=$benchmark_results_dir
 
+# now that the benchmark_run_dir directory exists, we can initialize the iterations file
+> ${benchmark_iterations}
+
 # For now, since we usually only ever run one iteration here, don't bother
 # collecting sysinfo data before and after, just do it after.
 # pbench-collect-sysinfo --group=$tool_group --dir=$benchmark_run_dir beg