`no_proxy_hosts` settings doesn't accept CIDR

### Description

I put the standard private network ranges into `no_proxy_hosts` (also tried it via ENV var `no_proxy`, and `NO_PROXY`):
```nix
no_proxy_hosts = [
"10.0.0.0/8"
"127.0.0.0/8"
"172.16.0.0/12"
"192.0.0.0/24"
"192.168.0.0/16"
"198.18.0.0/15"
];
```
(gets converted into YAML)

But the app service request toward `172.20.20.22` was failing with a timeout. I then added `172.20.20.22` into the `no_proxy_hosts` list and it started working.

Digging into the code a bit:

[here](https://github.com/element-hq/synapse/blob/develop/synapse/http/proxyagent.py#L276) the list of `no_proxy_hosts` get passed to `proxy_bypass_environment` which is a `urllib` function, which in `cpython` is defined [here](https://github.com/python/cpython/blob/abd19eddee20a7d05266f11f6042a84b04d29664/Lib/urllib/request.py#L1898). As you can see nowhere does it consider CIDR or IP ranges as documented by Synapse documentation:

> (array) List of hosts, IP addresses, or IP ranges in CIDR format which should not use the proxy. Synapse will directly connect to these hosts. For more details, see the [forward proxy documentation](https://element-hq.github.io/synapse/v1.138/setup/forward_proxy.html). There is no default for this option.

Either the documentation needs to be amended or CIDR + IP ranges implemented.

P.S.:

> It will not be used for:
>
> Application Services

and yet my `HS -> appservice` connection was failing

### Steps to reproduce

1. deploy synapse
2. deploy an appservice (`mautrix-discord` in my case)
3. deploy a forward proxy (`apache` in my case) and point synapse at it with `http_proxy` and `https_proxy`
4. exclude the private network range where the appservice resides at using `no_proxy_hosts`
5. observe `HS -> appservice` connection timeouts because it's going through the forward proxy

### Homeserver

matrix.redalder.org

### Synapse Version

v1.138

### Installation Method

Other (please mention below)

### Database

PostgreSQL

### Workers

Multiple workers

### Platform

- container
- `x86_64-linux-gnu`
- custom distro based on Nix: [https://github.com/nix-community/NixNG](NixNG)

### Configuration

_No response_

### Relevant log output

```shell
Oct 30 21:35:04 server container matrix-synapse[2474815]: 2025-10-30 22:35:04,422 - synapse.http.client - 467 - INFO - as-recoverer-26 - Error sending request to PUT http://172.20.20.22:29334/_matrix/app/v1/transactions/5990035: RequestTimedOutError 504: Timeout waiting for response from remote server
Oct 30 21:35:04 server container matrix-synapse[2474815]: 2025-10-30 22:35:04,423 - synapse.appservice.api - 412 - WARNING - as-recoverer-26 - push_bulk to http://172.20.20.22:29334 received code=504 msg=Timeout waiting for response from remote server
Oct 30 21:35:04 server container matrix-synapse[2474815]: 2025-10-30 22:35:04,423 - synapse.appservice.scheduler - 513 - INFO - as-recoverer-26 - Scheduling retries on mudiscord in 16.000000s
```

### Anything else that would be useful to know?

_No response_

Start in synapse/http/proxyagent.py around the call to urllib's proxy_bypass_environment, then reproduce the configuration with a CIDR entry and an appservice request. Compare the documented CIDR behavior with the current bypass handling; done means CIDR ranges in no_proxy_hosts reliably avoid the forward proxy, with the documentation or behavior aligned.

Written by the indexing model from the issue text.