In generated field name, an undescore is introduced after a number

### Summary

Underscore is added after a number in the field name, e.g. `c2s` becomes `c2_s`.

### Reproduction Steps

```proto
syntax = "proto3";

message Example {
string c2s = 1;
}
```

```bash
protoc --python_betterproto_out=lib example.proto
```

### Expected Results

Expecting `c2s`.

```python
# Generated by the protocol buffer compiler. DO NOT EDIT!
# sources: example.proto
# plugin: python-betterproto
# This file has been @generated

from dataclasses import dataclass

import betterproto

@dataclass(eq=False, repr=False)
class Example(betterproto.Message):
c2s: str = betterproto.string_field(1)
```

### Actual Results

Generates `c2_s` instead.

```python
# Generated by the protocol buffer compiler. DO NOT EDIT!
# sources: example.proto
# plugin: python-betterproto
# This file has been @generated

from dataclasses import dataclass

import betterproto

@dataclass(eq=False, repr=False)
class Example(betterproto.Message):
c2_s: str = betterproto.string_field(1)
```

### System Information

libprotoc 27.1
Python 3.8.20
Name: betterproto
Version: 2.0.0b7
Location: /Users/.../demo/.venv/lib/python3.8/site-packages
Requires: grpclib, python-dateutil, typing-extensions
Required-by: demo

### Checklist

- [x] I have searched the issues for duplicates.
- [x] I have shown the entire traceback, if possible.
- [x] I have verified this issue occurs on the latest prelease of betterproto which can be installed using `pip install -U --pre betterproto`, if possible.

Start by running the reported protoc command with the example.proto message and compare the generated field name with the expected output. Trace the generator entry point that converts protobuf field names to Python names; done means a field such as c2s remains c2s rather than becoming c2_s, with regression coverage if the project’s existing tests provide a suitable place.

Written by the indexing model from the issue text.