[add] Implement variable padding in number rounding

[initramfs]

This PR adds the ability to specify an integer value for the pad parameter within the round dictionary in order to "only" pad to the specific number of decimal digits.

To demonstrate with an example:

`mode: "places"`
#ztable(
    columns: 5,
    format: (
        (round: (precision: 4, pad: false)),
        (round: (precision: 4, pad: true)),
        (round: (precision: 4, pad: 1)),
        (round: (precision: 4, pad: 3)),
        (round: (precision: 4, pad: 100)),
    ),
    table.header[`pad: false`][`pad: true`][`pad: 1`][`pad: 3`][`pad` #sym.gt.eq `precision`],
    ..([3],) * 5,
    ..([3.1],) * 5,
    ..([3.14],) * 5,
    ..([3.141],) * 5,
    ..([3.1415],) * 5,
    ..([3.14159],) * 5
)

`mode: "figures"`
#ztable(
    columns: 5,
    format: (
        (round: (mode: "figures", precision: 4, pad: false)),
        (round: (mode: "figures", precision: 4, pad: true)),
        (round: (mode: "figures", precision: 4, pad: 1)),
        (round: (mode: "figures", precision: 4, pad: 3)),
        (round: (mode: "figures", precision: 4, pad: 4)),
    ),
    table.header[`pad: false`][`pad: true`][`pad: 1`][`pad: 3`][`pad: 4`],
    ..([0.3],) * 5,
    ..([0.31],) * 5,
    ..([0.314],) * 5,
    ..([0.3141],) * 5,
    ..([0.31415],) * 5,
    ..([0.314159],) * 5
)

would yield:

The main purpose of this is essentially to specify a minimum bound on the number of decimal digits, useful in cases like always displaying a decimal point (for integers), or to display to a fixed number of fractional digits on average but allowing more precise figures to exceed the digit count (by setting precision high and pad to the expected number of decimal digits on average).

---

This change remains backwards-compatible with the existing specification of pad: true or pad: false as it merely adds an integer overload ontop of the existing behaviour.

[Mc-Zen]

Hi @initramfs ,

thank you, this seems like a useful addition.

I think that in figures mode, it does not work correctly, though. I think that when the mode is set to "figures", the pad value should correspond to the number of significant figures. In the example you gave, it works by chance because there is no integer part but in general the count is wrong.

[initramfs]

I had already implement sigfig-based padding as seen here (but scrapped it for the initial PR):

Since this requires clarification of the semantics of pad (is it always digits or does it turn into sigfigs if mode: "figures").
This also incurs slightly extra complexity in round-or-pad() since we need to feed it the desired precision value as well.

I've added another commit that defines pad as being sigfig-based when in mode: "figures" since that seems to be the desired behaviour.

[Mc-Zen]

Looks good now! Can you add a few tests to tests/rounding/test.typ in the style of the last tests in the file?

And maybe we should mention that pad has no influence when mode: "uncertainty".

[initramfs]

Added tests and docs.

I amended the existing pad tests to expand on then rather than adding extra tests at the end, if that's okay with you.

[Mc-Zen]

Thank you! 🎉