If fiber's StreamRequestBody is enabled and request body compression is enabled, the request will fail.

## background

The request body is compressed using `Content-Encoding: br`.

- Disable StreamRequestBody: Request works normally
- Enabling StreamRequestBody: Request failed

## analyze

The server-side decompression layer was originally handled by fiber itself: fiber v3's DefaultReq.Body() ([req.go:150](https://github.com/gofiber/fiber/v3/blob/v3.3.0/req.go#L150)) would call BodyUnbrotliWithLimit to automatically decompress when it reads Content-Encoding. However, note the huma fiber v3 adapter [humafiber.go:106-113](https://github.com/danielgtaylor/huma/blob/v2.39.0/adapters/humafiber/humafiber.go#L106-L113):

```go
func (c *fiberWrapper) BodyReader() io.Reader {
if orig.App().Server().StreamRequestBody {
// Streaming is enabled, so send the reader.
return orig.Request().BodyStream() // ← raw compressed stream, bypassing Body() decompression.
}
return bytes.NewReader(orig.Body()) // ← fiber Body() will automatically decompress
}
```

After enabling StreamRequestBody, Huma reads the raw stream directly from Request().BodyStream(), bypassing the Content-Encoding decompression provided by fiber Body(); however, the Huma core does not handle HTTP layer Content-Encoding (it only recognizes base64 in the schema). Therefore, the Brotli stream enters the JSON parser as is → invalid character '\x1b' (0x1b is the first byte of the Brotli stream) → 400.

In other words, this is a real compatibility gap: huma (fiber v3) + StreamRequestBody will lose the decompression of the request body (Content-Encoding). It didn't cause this issue before because when BodyLimit was enabled or streams were not running, huma would decompress the request body via the orig.Body() branch.

This appears to be a regression of issue #1055.

Start with adapters/humafiber/humafiber.go around lines 106-113 and compare it with fiber v3's req.go around line 150. Reproduce a Brotli-compressed request with StreamRequestBody enabled, then verify that the request reaches Huma's JSON parser successfully and still behaves correctly without streaming.

Written by the indexing model from the issue text.