Extract a better property dict from TTF fonts

[jwiggins]

This fixes the font properties bugs pointed out by #391, but since it's happening in the refactored font code the issue remains open.

The core change here is to make get_ttf_prop_dict better.
Before:

In [3]: font = TTFont("/Library/Fonts/Times New Roman Bold.ttf")                

In [4]: get_ttf_prop_dict(font)                                                 
Out[4]: {'name': 'Times New Roman', 'sfnt4': 'Times New Roman Bold'}

After:

In [3]: font = TTFont("/Library/Fonts/Times New Roman Bold.ttf")                

In [4]: get_ttf_prop_dict(font)                                                 
Out[4]: 
{'copyright': '© 2006 The Monotype Corporation. All Rights Reserved.',
 'family': 'Times New Roman',
 'style': 'Bold',
 'unique_subfamily_id': 'Monotype:Times New Roman Bold:Version 5.01 (Microsoft)',
 'full_name': 'Times New Roman Bold',
 'version': 'Version 5.01.3x',
 'postscript_name': 'TimesNewRomanPS-BoldMT'}

Now that style is extracted, we can use it instead of the full_name of the font (which we used to call sfnt4) when checking the weight of the font.

As an added bonus, we're now using fonttools string decoding instead of the ad-hoc decoding implemented in #365

[jwiggins]

This change is integrating the face_index pieces that we're introduced by #605 to the refactored font parsing code.

[jwiggins]

I made this addition because there's a font named "Bodoni 72 Smallcaps Book" on macOS.

[jwiggins]

We used to use the full_name when checking the style and weight of a font.

[jwiggins]

We need this check to pick up older TTF/TTC files and newer OTF files (which use the Microsoft plaformID)

[rahulporuri]

LGTM

[jwiggins]

Okie dokie. Onward and upward I guess...