Keep original label names from C code

[afetisov]

I have added an Option<Rc<str>> field to the cfg::Label::FromC. This allows to preserve the names of blocks which are provided in the C source, potentially increasing the readability of transpiled code and making it easier to match original and transpiled code blocks. The label names are used as control flow variable values when --ddebug-labels is enabled. Unfortunately, Relooper really likes to dump those names in the _ branch of matches, bringing the automatic label names to the front. Still, there are some situations where the old labels can be traced in the transpiled code. This will also significantly increase readability if Relooper is changed to use labeled blocks & breaks instead of a match on variable: the C label names can be directly used as block label names.

I had to make Label non-Copy. This is not a painful transition, and arguably there was no reason to make it Copy in the first place, since it was likely to carry some non-Copy data.

[fw-immunant]

This looks good overall. My first thought was that it would be nice to keep Label: Copy, but that would mean accessing label_names in Label::pretty_print, and most callers of the latter don't have access to the TypedAstContext. So this approach seems sane and the implementation looks reasonable. The use of Rc here does seem worthwhile given how many .clone()s we end up introducing.

Most of the panics introduced seem to be only in places where we're relying on internal consistency of label_names with the set of AST locations where we expect names, so they shouldn't hurt overally reliability. The one exception is:

                    let label_name = from_value::<Rc<str>>(node.extras[0].clone())
                        .expect("unnamed label in C source code");

where we're relying on the serialized C AST having the structure we expect. This is the same pattern as all the adjacent code, so I'm satisfied with tests passing to validate it.

As far as improving output of translation goes, there are a few improvements in the output for our test cases, e.g. gotos/irreducible.c where this change now brings over the l1 label. 👍