Rename target_word_size

[sanxiyn]

At the moment, target_word_size in conditional compilation attributes actually refers to pointer size. Since word size (or register size) and pointer size are different concepts, it should be renamed.

[huonw]

Seems like target_pointer_size would be more appropriate.

[nagisa]

Do we want to do some transition period? I.e.

• Support both target_word_size and target_pointer_size
• Release a snapshot
• Remove the support for target_word size

[thestinger]

I doubt there are a significant number of users outside of the low-level standard libraries. Most of those (like big integers) are likely incorrect because they really do want the largest hardware integer type.

[nagisa]

@thestinger the problem here are not users outside, but the std itself. If we just rename the attribute to something else, the code will build with stage0 but not stage > 0 and were we to replace the use of attributes everywhere as well, compiler would not build with stage0.

[sfackler]

You don't necessarily need a period of time with support for both. You can either just stick both attributes on the types with one used at stage0 and one after, or use cfg_attr if you don't want the unused attribute warnings.

[codyps]

Was there a reason this didn't also adjust the json field name?