See previous commit.

Previously, when creating an object from inert inputs (eg:
"hb_font_create(hb_face_get_empty())") we returned the inert
empty object.  This is not helpful as there are legitimate
usecases to do that.

We now never return the inert object unless allocation failed.

Tests are revised to reflect.

For two reasons:

1. User can always call hb_buffer_pre_allocate() themselves, and

2. Now we do a pre_alloc in add_utfX anyway, so the total number of
reallocs is limited to a small number (~3) anyway.  This just makes the
API cleaner.

Remove hb_ft_get_font_funcs() as it cannot be used by the user anyway.

Add hb_ft_font_set_funcs().  Which will make the font internally use
FreeType.  That is, no need for the font to have created using the
hb-ft API.  Just create using hb_face_create()/hb_font_create() and
then call this on the font (after having set font scale).  This
internally creates an FT_Face and attached to the font.

We need this to set data on objects safely without worrying that some
other thread unsets it by setting it at the same time.

Exposes the non-atomicity of user_data opertaions at this time because
we call finish() while still locked and modifying the object.  In fact,
I'm surprised that it doesn't deadlock.  It should.

Revealed many bugs in the (untested and known buggy) user_data
support.