Bug#12093 "SP not found on second PS execution if another thread 
drops other SP in between" and
Bug#21294 "executing a prepared statement that executes a stored 
function which was recreat"

Stored functions are resolved at prepared statement prepare only.
If someone flushes the stored functions cache between prepare and
execute, execution fails.

The fix is to detect the situation of the cache flush and automatically
reprepare the prepared statement after it.

mysql-test/r/ps_ddl.result:
  Update results (Bug#12093 and Bug#21294, the test cases are already
  in the source tree).
mysql-test/r/ps_ddl1.result:
  Update results (Bug#12093 and Bug#21294, the test cases are already
  in the source tree).
mysql-test/r/sp-error.result:
  Update results (Bug#12093 and Bug#21294, the test cases are already
  in the source tree).
mysql-test/t/ps_ddl.test:
  Modify the test to not expect an error where there is no error
  any more (Bug#12093, Bug#21294).
mysql-test/t/ps_ddl1.test:
  Modify the test to not expect an error where there is no error
  any more (Bug#12093, Bug#21294).
mysql-test/t/sp-error.test:
  Modify the test to not expect an error where there is no error
  any more (Bug#12093, Bug#21294).
sql/sp_cache.cc:
  Implement sp_cache_version() -- returns the current version of 
  a stored routines cache.
sql/sp_cache.h:
  Declare sp_cache_version().
sql/sql_prepare.cc:
  Keep track of stored functions cache version, and invalidate
  the statement if it changed between prepared statement
  prepare and execute (and the statement actually uses stored routines).