Merge branch 'sql_stacktrace'

* sql_stacktrace:
  Tuple unpacking of stacktrace variables for niceness.
  Add preliminary support for stacktraces to see where SQL queries are coming from.

Added formatting of "GROUP BY" and "HAVING", which now show up in trunk queries
and could have shown up previously if manually patched into the QuerySet.query
instance.

Also indent joined tables a bit more underneath the "FROM" statement: they are
sub-statements of "FROM".

Signed-off-by: Rob Hudson <rob@cogit8.org>

We always start a new line for "FROM" now. Previous code was assuming MySQL
quoting of identifiers (MySQL is a bit non-standard in using backquotes) and
was assuming the last thing before the FROM would be a quoted identifier, which
wasn't always true, particularly when extra() is used on querysets.

Signed-off-by: Rob Hudson <rob@cogit8.org>

I'm wanting to test this a bit more before I merge it in fully.  Feedback welcome.

Fixed by moving the template panel's context_processor introspection to the
content method so this happens at the process_response time instead of at
process_request time.  Since context processors _were_ happening at the
process_response end, it was triggering the query
`user.get_and_delete_messages()` which was also getting triggered separately if
used in templates (i.e. the admin templates) resulting in duplicate queries
showing up in the toolbar.

the basic timings are shown for those that don't have the resource module, and 
finer grained timings are shown for those that do.

(some items were commented out after I was reminded that not all of the original BSD items are supported under Linux)