filter-repo: avoid aborting on a second --analyze run

Users may want to run --analyze both before and after filtering in
order to both find the big objects to remove and to verify they are
gone and the overall repository size and filenames are as expected.
As such, aborting and telling the user there's a previous analysis
directory in the way is annoying; just remove it instead.

Signed-off-by: Elijah Newren <newren@gmail.com>

git-filter-repo

@@ -16,6 +16,7 @@ import collections
 import fnmatch
 import os
 import re
+import shutil
 import StringIO
 import subprocess
 import sys
@@ -2475,8 +2476,7 @@ class RepoAnalyze(object):
       os.mkdir(results_tmp_dir)
     reportdir = os.path.join(results_tmp_dir, "analysis")
     if not args.force and os.path.isdir(reportdir):
-      raise SystemExit("Error: {} already exists; refusing to overwrite!".
-                       format(reportdir))
+      shutil.rmtree(reportdir)
     os.mkdir(reportdir)
 
     # Gather the data we need