Negative numbers skipped over by size/rem transform

[Tenga]

Description

When a value of a dimension token that would normally be transformed by the size/rem encounters a negative number, it's skipped and not transformed. size/px for example, doesn't seem to have this issue.

From what I can tell, this issue is due to a check in size/rem, that is specific to that transform.

Reproduction

Input file

{
  "typography": {
    "letterSpacing": {
      "$type": "dimension",
      "test1": {
        "$value": -1
      },
      "test2": {
        "$value": 1
      }
    }
  }
}

json/nested output with size/rem

{
  "typography": {
    "letterSpacing": {
      "test1": -1,
      "test2": "1rem"
    }
  }
}

json/nested output with size/px

{
  "typography": {
    "letterSpacing": {
      "test1": "-1px",
      "test2": "1px"
    }
  }
}

[jorenbroekema]

https://developer.mozilla.org/en-US/docs/Web/CSS/number#examples we could add +-e as valid number characters in the regex so it doesn't false-positively think there is already a unit.

That said, I think doing parseFloat removes units from stringified numbers already which kinda means we may not need to worry about doing a check whether the value already has a unit. It gets stripped anyways by parseFloat. This is why size/px works and isn't bugged. We should add plenty of tests though to ensure we don't make false assumptions about the parseFloat stripping units behavior