Emoji and abbreviations parser

[MihaZupan]

Fixes #296
Fewer memory allocations:

Building a new pipeline:

Method	Mean	Gen 0/1k Op	Gen 1/1k Op	Allocated Memory/Op
Markdig	12.05 us	19.1498	-	14.72 KB
Markdig_Advanced	42.11 us	45.1660	-	34.75 KB
Markdig_Advanced_Emoji	1,490.56 us	285.1563	130.8594	1502.8 KB
Markdig_Advanced_Emoji_Modified	1,499.65 us	298.8281	126.9531	1502.77 KB

(new) Method	Mean	Gen 0/1k Op	Gen 1/1k Op	Allocated Memory/Op
Markdig	12.04 us	19.1498	-	14.72 KB
Markdig_Advanced	41.97 us	45.1660	-	34.75 KB
Markdig_Advanced_Emoji	194.58 us	76.6602	15.3809	134.75 KB
Markdig_Advanced_Emoji_Modified	252.09 us	85.4492	18.5547	162.05 KB

Where Modified forces the lazy-init of dictionary properties.

Parsing speed for emojis and abbreviations is about the same (~10% faster for emojis),
as the dataset could be considered the worst-case for a prefix tree of this type (every input starts with the same character).

[xoofx]

Thanks, great perf improvement!