Fix page size for aarch64-apple-darwin in test

[youknowone]

Page size is not always 4k

[lhecker]

I don't think this is the right solution. IMO either we should hardcode it to 16Ki (or 64Ki or more) because the test doesn't actually care about the page size, or we should add a function to sys that returns the OS page size.

What do you think?

[youknowone]

Because the value is only used in test, I selected hard-coding big value.

page_size crate is used by many crates for this purpose, but this project seems to keep the dependency as small as possible.

[lhecker]

LGTM!

[lhecker]

[...] but this project seems to keep the dependency as small as possible.

There are various reasons for this, but as I mentioned elsewhere, I'm trying to keep the editor fast and small. Dependencies don't always compose well. Someone asked for instance why I don't use the memchr crate, and that's because our memchr is faster for our specific use case. Someone else asked why I don't use bumpalo, and that's because this wouldn't tightly integrate with the scratch_arena pattern that I like. unicode-rs is problematic because it's too large. Same for regex. And so on...

For page_size however, I'd not use it because this project gets deployed to almost a billion devices worldwide and I don't want to take any risks using dependencies that aren't extraordinarily reputable.

Personally, I also like it because I can learn new tricks this way. What better way is there to grow as an engineer than to challenge oneself?

[youknowone]

In this case, page_size can be added to dev-dependency. Then it will be run only in test but never will be deployed to users.