High compile time of crates using sqlparser(codegen phase) - any way to reduce generated code?​​

In my project, some crates have very long compilation times (e.g., 40-50 seconds, with the codegen phase accounting for 70%-90% of that time). I suspect these crates are slow to compile due to their dependency on sqlparser.

I analyzed the compilation of one such crate using cargo llvm-lines. The top functions by lines are listed below:

```
Lines Copies Function name
----- ------ -------------
2180708 31532 (TOTAL)
220326 (10.1%, 10.1%) 21 (0.1%, 0.1%) ::visit
163970 (7.5%, 17.6%) 1995 (6.3%, 6.4%) as sqlparser::ast::visitor::VisitMut>::visit
117960 (5.4%, 23.0%) 3654 (11.6%, 18.0%) as sqlparser::ast::visitor::VisitMut>::visit
76507 (3.5%, 26.5%) 21 (0.1%, 18.0%) ::visit
```

As shown, the visitor trait implementations for the AST types generate a significant amount of code. Is there any way to reduce the code generation footprint of sqlparser?

Start by reproducing the reported compilation profile with cargo llvm-lines and inspect the generated VisitMut implementations for the AST types named in the report. Compare the codegen footprint and compilation time before and after any approach; done means a demonstrated reduction in generated code or compile time without breaking visitor behavior.

Written by the indexing model from the issue text.