Add coverage for U+0326, U+0327, and U+0328

[MariannaPaszkowska]

Hi, I'd like to report that those glyphs are not included in any subset definitions:
“◌̦” (U+0326)
“◌̧” (U+0327)
“◌̨” (U+0328)
I think it is needed to include those somewhere, perhaps in the latin-ext_unique-glyphs.nam? Would that be possible?

Those are always getting subset out and shaperglot identifies those as used by the following languages:
“◌̦” (U+0326):

Aromanian [rup_Latn], Gagauz [gag_Latn], Marshallese [mh_Latn], 
Prussian [prg_Latn], Romanian [ro_Latn], Turkish [tr_Latn], Turkmen (Latin) [tk_Latn]

“◌̧” (U+0327):

Albanian [sq_Latn], Arabic, Chadian Spoken [shu_Latn],
Azerbaijani [az_Latn], Catalan [ca_Latn], Colognian [ksh_Latn],
Dii [dur_Latn], Doyayo [dow_Latn], English [en_Latn], Ewe [ee_Latn],
French [fr_Latn], Friulian [fur_Latn], Gagauz [gag_Latn], Gbaya-
Mbodomo [gmm_Latn], Gheg Albanian [aln_Latn], Kabuverdianu [kea_Latn],
Kako [kkj_Latn], Karang [kzr_Latn], Kirmanjki [kiu_Latn],
Koalib [kib_Latn], Kurdish (Latin) [ku_Latn], Latgalian [ltg_Latn],
Latvian [lv_Latn], Laz (Latin) [lzz_Latn], Makaa [mcp_Latn],
Mandjak [mfv_Latn], Mankanya [knf_Latn], Manx [gv_Latn],
Marshallese [mh_Latn], Mirandese [mwl_Latn], Muslim Tat, Latin [ttt_Latn],
Occitan [oc_Latn], Pana, Central African Republic [pnz_Latn],
Piaroa [pid_Latn], Picard [pcd_Latn], Portuguese [pt_Latn],
Prussian [prg_Latn], Romanian [ro_Latn], Saho [ssy_Latn],
Sardinian [sc_Latn], Sar [mwm_Latn], Sassarese Sardinian [sdc_Latn],
Sãotomense [cri_Latn], Talysh (Latin) [tly_Latn],
Tsakhur (Latin) [tkr_Latn], Turkish [tr_Latn], Turkmen (Latin) [tk_Latn],
Venetian [vec_Latn], Vute [vut_Latn], Walloon [wa_Latn], Walser [wae_Latn],
Zaza [zza_Latn], Zulu [zu_Latn]

“◌̨” (U+0328):

Chipewyan [chp_Latn], Dogrib [dgr_Latn], Elfdalian [ovd_Latn],
Gwichʼin [gwi_Latn], Han [haa_Latn], Icelandic [is_Latn],
Kashubian [csb_Latn], Kaska [kkz_Latn], Lithuanian [lt_Latn],
Marshallese [mh_Latn], Navajo [nv_Latn], Palauan [pau_Latn],
Polish [pl_Latn]

[simoncozens]

Unfortunately latin-ext will not work.

Let's say we put the combining characters into latin-ext. If we want to write ə̧ for the Dii language, ə is in latin-ext and ◌̧ is in latin-ext. Both base and mark are served from the same subfont and all is well. But the minute we want to write m̧ (for the Marshallese language), we have a problem: m is in latin and ◌̧ is in latin-ext, base and mark are now served from two different subfonts and will not combine. (This is further explored in #6) But of course if we move ◌̧ back to latin, we break latin-ext combinations.

What we want is for combining characters to be in latin and latin-ext and vietnamese and any other script which may use these Common Script codepoints. But the way the subsetter is currently designed, codepoints need to live in a single subset, so this option is not available either. (It's also not totally clear that browsers will do the right thing when codepoints are supplied in multiple subfonts, but I have no idea what the current state of the evidence is on this.)

The long term fix is Incremental Font Transfer, but that is still some way away.