Embed errors in the AST instead of raising

Currently, when `ppx_fields_conv` encounters an error, it uses the `raise_errorf` function to raise a located error.

The exception is caught by `ppxlib`, which in this case:

- Catch the error,
- stops the rewriting process
- add the error (as a `[%%%ocaml.error ...]` extension node) to the _last valid ast_
- Use the resulting AST

The interruption of the rewriting is quite bad for the user experience! The implication for the users are:

- Since `ppx_fields_conv` runs at the "context-free" phase, the "last valid AST" is before the context-free phase. So, no other derivers/extenders get run, which generates a lot of noise in the errors (such as "uninterpreted extensions" or "unbound identifiers")
- Only one (meaningful) error from your PPX is reported at a time.
Example

For instance:
```ocaml
type invalid1 = int [@@deriving fields]

type invalid2 = int [@@deriving fields]

type valid = { i : int } [@@deriving fields]
let _ = i
```
would report several errors:
- `Unsupported use of fields (you can only use it on records` for `invalid1`: the right error
- 'Unbound value i`

The error for `invalid2` is not shown, since the rewriting is cancelled when the exception is raised.

You can find more information about error reporting in PPXs in [this section](https://ocaml.org/p/ppxlib/latest/doc/good-practices.html#handling_errors) of the ppxlib manual.

:question: Would you be willing to accept contributions to this issue? I'm considering assigning its resolution as part of an [outreachy](https://www.outreachy.org/) internship: see more information [here](https://github.com/ocaml-ppx/ppxlib/issues/389).

Start by reading the ppxlib manual's error-handling section and inspect how ppx_fields_conv currently uses raise_errorf. Use the invalid1/invalid2/valid example as the behavioral check: errors from both invalid declarations should be preserved, later derivers should still run, and unrelated noise should be avoided.

Written by the indexing model from the issue text.