pretty print source comparison failures are not fatal

[tinco]

Since 7948afd the compare_source check no longer panics on failure. It is obvious in the commit that the panic invocation is removed from the check, and no other changes are made in the function to remedy it.

It can be verified by adding a new pretty test case like:

// pp-exact
// The next line should not be expanded
mod issue_12590_b;

The resulting pretty printed file is slightly different, the comparison fails and the message is logged, but the test continues into the type check, which fails for a different reason and then panics. If you fix the type check (by adding the module and fn main) then the whole test passes and the error message is swallowed even though the source comparison is still not matching.