Custom Output Naming

### Feature: Custom Output Naming

I've found `docling` to be great, but simply using the basename of whatever URI we're hitting isn't always ideal.
I'd like to be able to have `docling` save documents according to some information within the document, or to include additional parts of the URI.

1. URI-based Enhancements
* Additional Folders/Namespaces
* domain.tld/2025-10-01/post.html -> `2025-10-01_post.md` vs `post.md`
* "Always" and "When Name < N Characters in Length"
* Domain Prefixing
* domain_one.tld/some/path/How_to_Win.html -> `domain_one_tld_How_to_Win.md`
* domain_two.tld/some/path/How_to_Win.html -> `domain_two_tld_How_to_Win.md`
* Domain Bucketing
* `docling --output=~/Documents domain.tld/2025-10-01/post.html`
* --> `~/Documents/domain.tld/post.md`
* Preserving "Extension"
* Define a list of common/known web extension (ie: .html, .php)
* Only trim strings for these values, not "everything after the last dot".
1. Metadata-based Enhancements
* Use `` for naming instead of file basename.
* Fallback on URI-based naming if not available
* LLM naming
* Allow the LLM to come up with a title for the document.
1. Date-based Enhancements
* Allow custom `date` timestamps to be used when naming the output.
1. Manual
* Expose `--filename`
* Relative to `--output` or CWD
* Absolute and bypass `--output`

We should also have the option of normalizing/sanitizing this output string:

* Special Character -> `_`
* Space and `-` -> `_`
* Emoji / Extended Unicode -> `_`

#### Use Case

`docling` is great, until you grab a document and it's named `aarch64.md`, and then you're afraid you might blow it away with another document with a similar name. I am looking for a way to help prevent name clashing, while not asking for a "versioning" or "backup" feature (though, I do think that would also be helpful). While this could be used to help with clashing, the primary goal is to allow better/custom naming of file outputs.

I checked `docling --help`, but I did not see anything that would help me accomplish any of this.
I primarily use `docling` via it's CLI implementation.

#### Examples

1. `https://csrc.nist.gov/pubs/fips/203/final`
* I'd like to be able to grab this as `203_final`
1. `https://doi.org/10.6028/NIST.FIPS.203`
* Here, the file is downloaded as `NIST.FIPS`, which means they clash
1. `https://xkfan.github.io/cross-toolchain/aarch64.html`
* This doesn't result in a valid toolchain, but it is an example document.

### Alternatives
I could fire off and request the page data myself, extract `<title>` and then rename everything... but...
1. I'd need to know the file(s) generated by `docling`
1. It'd be hairy and hacky
1. `docling` is already doing a great job with what it does, see if it makes sense to implement within