shard audit exaggerates reports of overlapping shard ranges

When sharder audits a root container and finds overlapping ranges, there is a tendency for each overlap to be logged more than once. For, example the following might be logged:

Audit failed for root var/folders/n4/nf8krx110f39llj82fbw_xjm0000gp/T/tmpFbb92K/sda/containers/0/f33/4a8a08f09d37b73795649038408b5f33/4a8a08f09d37b73795649038408b5f33.db (a/c): overlapping ranges in state created: s-z k-t, overlapping ranges in state created: k-t s-z

This can be illustrated by changing a unit test to assert that both variations of the same overlap are logged:

diff --git a/test/unit/container/test_sharder.py b/test/unit/container/test_sharder.py
index 22cd86797..7df1d145b 100644
--- a/test/unit/container/test_sharder.py
+++ b/test/unit/container/test_sharder.py
@@ -4397,6 +4397,9 @@ class TestSharder(BaseTestSharder):
             self.assertIn(
                 'overlapping ranges in state %s: k-t s-z' % state_text,
                 line)
+ self.assertIn(
+ 'overlapping ranges in state %s: s-z k-t' % state_text,
+ line)

         expected_stats = {'attempted': 1, 'success': 0, 'failure': 1}
         shard_bounds = (('a', 'j'), ('k', 't'), ('s', 'z'))

(based on commit 257041aff)