*improved draw sprite so that it works within a group

*patch to add [group] data structure and supporting fixes (work in progress with lingering bug of being unable to display data structure arrays properly, see tiger.pd demo patch in data structures)

*partial backport of a fix to allow subpatches to be renamed with no arguments without losing their inner content (see g_text setto call). Thanks to Jonathan Wilkes for the bug report.

*improved comment resize bar (made it into a dashed box)
*improved invalid object outline (box) into a dashed box and made such boxes solid when being edited to distinguish them from comments.

*restructured pdsend messages used by sys_gui object so that they are forward and backward compatible.

*fixed pd~ regression where instantiation within another instance of pd~ failed to start due to expectation of the new -unique flag without which all patches are opened within the existing instance.

- nested plots display properly
- plot is vis'd and unvis'd properly when deleting plot or struct
- commented out ugly plot_displace hack-- multiple plots should work just fine
- separated out garray dependencies into class "old_plot".  This way the changes I made don't affect garrays.  _Very_ ugly hack, but it was pretty simple to implement and will be easy to remove later.
- no crashes when non-existent arrays are entered into a [struct] that has scalars hanging around
- template_cancreate ported from 0.45
- added a function to get the offending struct (to use with "Find last error").  Not using it yet, though.
+ For creating scalars using an object box:
- scalar must have initial "float x float y" fields, in that order.  This means you can't accidentally type the name of the hidden garray structs and create a scalar.  (Nor any other structs that only have a "y" field, etc.)
- broken object box returned if the struct has a bogus array template.  (Although currently I'm just refusing to create such a struct.)

improved installer to avoid install of Gem libraries when doing incremental install (should significantly shorten incremental installs)