Always return WT_TRY_SALVAGE for corrupted log

As a followup to WT-4344, there are some cases for which a damaged/missing/truncated log file will give an error other than WT_TRY_SALVAGE.  The easiest way to see this is to edit test_txn19.py and in this fragment:

         if expect_fail:

             self.check_file_contains_one_of(errfile,

                 ['/log file.*corrupted/',

                 'WT_TRY_SALVAGE: database corruption detected'])

remove the log file.*corrupted line.  This produces failures for cases where are log file is either zero length or begins with garbage.  For those cases we don't get WT_TRY_SALVAGE, but do get a log file...corrupted message. Presumably having an accurate return value will allow MongoDB to suggest the best course of action for recovery.