feat: dont do needless search when we reach max depth

[sigmaSd]

fix #2565

I can improve the code, if in general it seems reasonable,
seems to work in my testing

[sigmaSd]

Btw this gets it down to 3 syscalls per entry when max depth is reached

openat(AT_FDCWD, "/home/mrcool/dev/rust/nanoserde", O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_DIRECTORY) = 4
newfstatat(4, "", {st_mode=S_IFDIR|0755, st_size=4096, ...}, AT_EMPTY_PATH) = 0
close(4)                                = 0
openat(AT_FDCWD, "/home/mrcool/dev/rust/message-io", O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_DIRECTORY) = 4
newfstatat(4, "", {st_mode=S_IFDIR|0755, st_size=4096, ...}, AT_EMPTY_PATH) = 0
close(4)                                = 0
openat(AT_FDCWD, "/home/mrcool/dev/rust/mun_example", O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_DIRECTORY) = 4
newfstatat(4, "", {st_mode=S_IFDIR|0755, st_size=4096, ...}, AT_EMPTY_PATH) = 0
close(4)                                = 0
openat(AT_FDCWD, "/home/mrcool/dev/rust/progress", O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_DIRECTORY) = 4
newfstatat(4, "", {st_mode=S_IFDIR|0755, st_size=4096, ...}, AT_EMPTY_PATH) = 0

But I imagine even those are not needed, but I guess its probably negligible

[BurntSushi]

Thank you, but I'm going to pass on this for now. I hope to do an overhaul of the ignore crate soon, and I'll see if it makes sense to take max depth into account at a more fundamental level. I'm not quite comfortable with this change as-is because I think it's a little too hacky. (Which isn't your fault. The crate's internals are an awful mess and I barely understand it any more.)