Support for numerals in the data and detection

[MrBrezina]

We thought this approach would work well:

• define numerals for the script,
• languages automatically inherit this definition, but
• can optionally override it.

Alternatively, we could use cross-language inheritance as it is already used for some orthographies. To avoid introducing a new linking logic.

Design requirements/alternates should mention any language-specific preferences, e.g. Urdu preference differs from the default Persian numerals to which these are usually mapped. I am not sure this should be a shaping check that fails detection as some fonts may be designed with Urdu numerals as default.

From the top of my head, I do not see the case for auxiliary numerals. Maybe there is some? Even in the case of Devanagari, which has its own numerals, I would list also include the Latin-script numerals (I will not call these Arabic!) as both sets are used frequently.

Language detection should fail if the numerals are defined, but not included in the font.