Clarify longest match behavior of the grammar

[msprotz]

https://doc.rust-lang.org/reference/expressions.html claims as binds tighter than than + or <

Indeed, with +: let b = 0 as u32 + 0 as u32 parses correctly,

However, with <: let b = 0 as u32 < 0 as u32 gives a syntax error.

I suspect this is due to a conflict in the grammar that looks ahead at the token < and sees the beginning of a type application (in types), but if I'm reading just https://doc.rust-lang.org/reference/expressions.html#r-expr.precedence there's not enough information for me to write a correct Rust syntax printer.

[ehuss]

Generally, though it is not fully documented, the productions are greedy in a sense that they will attempt to parse the longest possible set of tokens for a rule (even if that would lead to an error). So, this isn't exactly an operator precedence issue. It is TypePath is greedily taking tokens like < 0 and giving up at some point (probably the as) with an error.

[msprotz]

And maybe explain that the grammar won't backtrack (past N tokens, I presume) -- with backtracking one could possibly parse this expression, but it would be expensive.