Fixed bugs in calculating text size

[radarhere]

Resolves #3745

[hugovk]

Please include a summary in PRs of what is changed and why. It makes them easier to review. Thanks!

[hugovk]

What's the reason for moving these declarations before use?

[radarhere]

x_advanced is new, not moved. I'm happy to move the declarations down if you like, I'm just going with the convention established in the code - I moved the offset declaration up because it is no longer inside a conditional block.

[radarhere]

With English characters, each subsequent character ends horizontally after all of the previous ones. Pillow has a variable x, and it moves it forward with each character. Pillow presumes the last x is the maximum x value.

From the issue, there are two characters - the backwards S at the bottom, and the line at the top. The backwards S ends horizontally after the line at the top - the horizontal clipping occurs because Pillow stops the size after the line at the top. So instead, check each character to see if it has the maximum x value.

The first diagram at https://www.freetype.org/freetype2/docs/tutorial/step2.html is helpful in understanding the glyph properties.

With regards to changing the vertical clipping, I presume that Pillow is not correctly processing the fact that the line at the top does not start at the baseline - there is a gap. Flipping the signs when applying the offset fixes this problem, with the only affect on existing tests being that a text string is moved vertically up.

Let me know if anything is still unclear.

[hugovk]

Please resolve the merge conflicts and let's merge.