Fixed dotted circle deletion in incorrectly ordered Om sequence

[iwsfutcmd]

fixes #32, as well as the problem noted in #1 (comment)

Note: Noto Sans Javanese before this fix allowed two sequences for the Javanese Om:

the correct:

ꦎ A98E JAVANESE LETTER O
ꦴ A9B4 JAVANESE VOWEL SIGN TARUNG (long -aa vowel mark)
ꦀ A980 JAVANESE SIGN PANYANGGA (candrabindu)

and the incorrect:

ꦎ A98E JAVANESE LETTER O
ꦀ A980 JAVANESE SIGN PANYANGGA
ꦴ A9B4 JAVANESE VOWEL SIGN TARUNG

Because the second sequence is considered incorrect by Unicode (the correct order for Javanese should be in phonetic order), a dotted circle is automatically inserted between the PANYANGGA and TARUNG, as PANYANGGA should always occur after any vowel marks in memory. To get around this, the shaping rules went back and deleted the dotted circle. This had the side effect of making the sequence:

ꦎ A98E JAVANESE LETTER O
ꦀ A980 JAVANESE SIGN PANYANGGA
◌ 25CC DOTTED CIRCLE
ꦴ A9B4 JAVANESE VOWEL SIGN TARUNG

also display as the om_javanese glyph.

So to prevent the sequence with 25CC from rendering as om_javanese, we need to also remove the functionality that shapes the incorrect order as om_javanese.

This should be fine, and preferable

Based on my research, the incorrect sequence is practically not found in the wild—the only usages that comes up on Google are a bot-driven site that attempts to read PDFs, and the first result is actually this repo.

On the other hand, the correct sequence has 6 pages of legitimate usage, including Wiktionary, Wikipedia, and many social media posts.

Fixing Noto Sans Javanese to not allow the incorrect order to render correctly would be helpful for not propagating the incorrect order, in addition to solving this disappearing dotted circle issue.

[simoncozens]

Thank you!