Prevent attempts at nesting character sets in regex filter

[JoshuaS3]

Not sure how I didn't catch this when testing the initial regex feature a couple years ago. Basically, nesting character sets is impossible in Python's regex, which is what my filter code tried to do because it wasn't checking for IN tokens, so those confusable glyphs would be discarded by the regex compiler (thankfully it doesn't outright throw an error). This PR adds that check and performs confusable insertion in place for those sets. Also cleans up some of the code in that area.

Example of the issue: h[ea]lp would effectively be converted into this, which is invalid regex:

[hｈℎ𝐡𝒉𝒽𝓱𝔥𝕙𝖍𝗁𝗵𝘩𝙝𝚑һհᏂ]**[**[e℮ｅℯⅇ𝐞𝑒𝒆𝓮𝔢𝕖𝖊𝖾𝗲𝘦𝙚𝚎ꬲеҽ][a⍺ａ𝐚𝑎𝒂𝒶𝓪𝔞𝕒𝖆𝖺𝗮𝘢𝙖𝚊ɑα𝛂𝛼𝜶𝝰𝞪а]**]**[l𝟏𝟙𝟣𝟭𝟷IＩⅠℐℑ𝐈𝐼𝑰𝓘𝕀𝕴𝖨𝗜𝘐𝙄𝙸Ɩｌⅼℓ𝐥𝑙𝒍𝓁𝓵𝔩𝕝𝖑𝗅𝗹𝘭𝙡𝚕ǀΙ𝚰𝛪𝜤𝝞𝞘ⲒІӀⵏᛁꓲ𖼨𐊊𐌉][p⍴ｐ𝐩𝑝𝒑𝓅𝓹𝔭𝕡𝖕𝗉𝗽𝘱𝙥𝚙ρϱ𝛒𝛠𝜌𝜚𝝆𝝔𝞀𝞎𝞺𝟈ⲣр]

instead of this, which is valid and desired:

[hｈℎ𝐡𝒉𝒽𝓱𝔥𝕙𝖍𝗁𝗵𝘩𝙝𝚑һհᏂ]**[**e℮ｅℯⅇ𝐞𝑒𝒆𝓮𝔢𝕖𝖊𝖾𝗲𝘦𝙚𝚎ꬲеҽa⍺ａ𝐚𝑎𝒂𝒶𝓪𝔞𝕒𝖆𝖺𝗮𝘢𝙖𝚊ɑα𝛂𝛼𝜶𝝰𝞪а**]**[l𝟏𝟙𝟣𝟭𝟷IＩⅠℐℑ𝐈𝐼𝑰𝓘𝕀𝕴𝖨𝗜𝘐𝙄𝙸Ɩｌⅼℓ𝐥𝑙𝒍𝓁𝓵𝔩𝕝𝖑𝗅𝗹𝘭𝙡𝚕ǀΙ𝚰𝛪𝜤𝝞𝞘ⲒІӀⵏᛁꓲ𖼨𐊊𐌉][p⍴ｐ𝐩𝑝𝒑𝓅𝓹𝔭𝕡𝖕𝗉𝗽𝘱𝙥𝚙ρϱ𝛒𝛠𝜌𝜚𝝆𝝔𝞀𝞎𝞺𝟈ⲣр]

(** only added for emphasis)