Reading the discussion of this feature, I gained the mistaken impression that neither `set` nor `get` blocked, and thus calling `get` immediately after `set` was not guaranteed to succeed. It turns out that `set` *does* block, guaranteeing that the cell contains a value once `set` returns. This change updates the documentation to state that explicitly.

Almost all safety comments are of the form `// SAFETY:`,
so normalize the rest and fix a few of them that should
have been a `/// # Safety` section instead.

Furthermore, make `tidy` only allow the uppercase form. While
currently `tidy` only checks `core`, it is a good idea to prevent
`core` from drifting to non-uppercase comments, so that later
we can start checking `alloc` etc. too.
Signed-off-by: NMiguel Ojeda <ojeda@kernel.org>

Fixes #76367.

Also doing fmt inplace as requested.

I think this communicates the intent better, and is shorter anyway.

They were carried over from once_cell crate, but they are not entirely
correct (as miri now supports more things), and we don't run miri
tests for std, so let's just remove them.

Maybe one day we'll run miri in std, but then we can just re-install
these attributes.

See the failed build in

https://github.com/rust-lang/rust/pull/75555#issuecomment-675016718

for an example where we need this in real life

The logic here is the same as for Send&Sync impls.

Spinning multiple threads in this test causes a deadlock in
SGX where thread scheduling is not preemptive.

This commit refactors the initial implementation to fit into std and
makes some other changes:

- use MaybeUninit internally in SyncOnceCell
- correctly impl Drop for lazy::Once
- port Lazy::take from once_cell from: https://github.com/matklad/once_cell/pull/100Co-Authored-By: NPaul Dicker <pitdicker@users.noreply.github.com>