Primitive bit validity and padding guarantees

[joshlf]

Also fix a few typos.

Resolves #1291

[ehuss]

Thanks for the PR! Can you say more about why it seems that this should be part of the layout chapter? My perspective is that the layout chapter is relevant to the alignment and padding of fields within a nominal type (and a few other circumstances), but the internal representation of a type is not relevant here. The types chapters defines what the valid bit patterns are (and by extension, the UB chapter explains invalid values and safety).

I'm reluctant to define this multiple times, or go into details about what kind of transmutes are sound within the layout chapter.

[jswrenn]

Suggested change

	For each primitive type, `T`, in the preceding table other than `char`, any
	For each primitive type, `T`, in the preceding table other than `char` and `bool`, any

[jswrenn]

Suggested change

	`char`), `T` contains no padding or otherwise uninitialized bytes. In other
	`char` and `bool`), `T` contains no padding or otherwise uninitialized bytes. In other

[jswrenn]

Suggested change

	#### `bool` Bit Validity
	A `bool`'s numerical value is guaranteed to be either 0x00 or 0x01,
	It is undefined behavior to construct a `bool` with a value outside
	this range. See the [`bool` docs][bool-docs] for more information.

[jswrenn]

Suggested change

	[char-docs]: ../std/primitive.char.html
	[bool-docs]: ../std/primitive.bool.html
	[char-docs]: ../std/primitive.char.html

[jswrenn]

Hm, this phrasing struck me as potentially confusing, since "x86" is often used as a shorthand to refer to "x86-64". (I did a brief double take, at least.) Per Wikipedia's chronology, "IA-32" might be a more appropriate term:

Suggested change

	platform-specific behavior. In particular, on x86, `u64` and `f64` are only
	platform-specific behavior. In particular, on IA-32, `u64` and `f64` are only

...but I think you're referring to target_arch value "x86" (which is distinct from x86_64). Rust inherits this shorthand from LLVM:

    x86,            // X86: i[3-9]86

...so you could instead disambiguate this by writing:

Suggested change

	platform-specific behavior. In particular, on x86, `u64` and `f64` are only
	platform-specific behavior. In particular, on `target_arch = x86` (which has a word size of 32 bits), `u64` and `f64` are only

[joshlf]

@ehuss I didn't realize that validity was documented elsewhere! I've put up #1293, which adds a bit of documentation to the pages in the "Types" section. Thanks for the feedback!