[bug] DynaconfFormatError (circular reference) when applying @jinja to alter the same value

**Describe the bug**
I apply the following code to make sure that a value is an integer even if defined as string:
`MAIN__STARTYEAR = "@int @jinja {{this.main.startyear | int}}"`
This is part of an toml-file defined via the includes-parameter of `Dynaconf`

Before version 3.3.0 this went well, now it raises a DynaconfFormatError (Circular reference detected in lazy formatting. A value is referencing itself directly or indirectly.) in nodes.py:734

**To Reproduce**
Steps to reproduce the behavior:

1. Having the following folder structure

Project structure

```bash
# /dynaconf_circular/
```

2. Having the following config files:

Config files

**/dynaconf_circular/settings.toml**
```toml
[main]
startyear = "2015"
```
and

**/dynaconf_circular/config_post.toml**
```bash
MAIN__STARTYEAR = "@int @jinja {{this.main.startyear | int}}"
```

3. Having the following app code:

Code

**/dynaconf_circular/config.py**
```python
from dynaconf import Dynaconf

if __name__ == '__main__':
settings_dyna = Dynaconf(
merge_enabled=True,
environments=False,
settings_files=["settings.toml"],
loaders=[
"dynaconf.loaders.toml_loader",
],
includes=["config_post.toml"],
)
print(settings_dyna.main.startyear)
```

4. Executing under the following environment

Execution

```bash

$ python config.py
```

**Expected behavior**
Printing "2015"

**Environment (please complete the following information):**
- OS: Windows 11
- Dynaconf Version 3.3.0

**Additional context**
For the DynaconfFormatError, it would be good to include more context (i.e. the value that is circularly referenced) in the error message!

Reproduce the circular-reference failure with settings.toml, config_post.toml, and config.py, then inspect nodes.py around line 734 to trace the lazy-formatting behavior. Done means the example prints 2015 as expected and the DynaconfFormatError provides more context about the circularly referenced value; an open pull request is already linked.

Written by the indexing model from the issue text.