Add font size warning to TextSection

[hate]

Objective

• TextSections are unable to display their text value if the font_size is less than or equal to 0, thus @ickshonpe suggested we emit a warning to the console whenever this occurs.
• Fixes Negative font_size warning #9655

Solution

• Emit a warning if we detect that the font_size is less than or equal to 0.
• Add bevy_log as a dependency to bevy_text to be able to access the warn! macro.

[hate]

Similarly, bevy_text::TextSection::from_styleshould emit a warning as well, but I am unsure how to call the warn! macro since the method signature is const

[alice-i-cecile]

Good work :) We could make from_section no longer const, although that would be a bit sad.

[KingTheSim]

Shouldn't the function not attempt to create a TextSection if the font size is invalid?

[alice-i-cecile]

That TextSection could later be updated to be valid though. Our options here are:

• silently fail
• warn
• return a Result (which will almost always get unwrapped)
• panic

Because the problem is repairable and cosmetic, I think that warning is the least bad option, although I'm open to being convinced.

[KingTheSim]

Silent fail won't give us information and Result always get's unwrapped, which doesn't add value.
If getting an invalid font size doesn't cause a crash/break down the line, I agree with the warning. We can later set a default value if that's ok. If it does break somewhere, panic with explanation is alright.

[ickshonpe]

That TextSection could later be updated to be valid though.

When I posted the issue my thinking was that the warning would be added somewhere to the text systems rather than to TextSection itself because of the inverse of this, this implementation won't pick up if the text is updated later to an invalid state.
And there would be a flag to make sure the warning isn't spammed, similar to FontAtlasWarning.

Warning on creation of TextSections didn't occur to me and has advantages too. It covers 99% of cases, most users will never change the font size after creation, and the implementation is far simpler.

[ravenclaw900]

While we're at it, would it also make sense to check if the number is_finite (not NaN, -Inf, or Inf)? I would agree that this is an obvious bug in the calling code, but so is passing in a negative, therefore I think this check would be reasonable.