Add parsing of PDF titles

This commit attempts to parse the title metadata in a PDF file if it
exists, otherwise it fallbacks to printing only the media type and
size.

We use the `pdf` crate to parse all of the response body (for some
reason metadata and table of contents are kept at the end of PDFs) and
then ask the PDF for the title that may be defined in the "[info
dictionary]".

[info dictionary]: https://www.adobe.com/content/dam/acom/en/devnet/pdf/pdfs/pdf_reference_archives/PDFReference.pdf

Author: Jassob

Cargo.lock

@@ -51,6 +51,7 @@ stderrlog = "0.5.1"
 atty = "0.2.14"
 scraper = { version = "0.12.0", default-features = false, features = [] }
 phf = "0.7.24"
+pdf = "0.7.1"
 
 [dependencies.image]
 version = "0.22.5"

Cargo.toml

@@ -199,6 +199,18 @@ pub fn get_title(resp: &mut Response, rtd: &Rtd, dump: bool) -> Result<String, E
         trace!("[{}] {}", k, v.to_str().unwrap());
     });
 
+    if let Some(mime) = content_type.clone() {
+        if mime == mime::APPLICATION_PDF {
+            let mut body = Vec::new();
+            if let Err(err) = resp.read_to_end(&mut body) {
+                return Err(format_err!("failed to read pdf body: {}", err));
+            }
+            return get_pdf_metadata(&body)
+                .or_else(|| get_mime(&rtd, &mime, &size))
+                .ok_or(format_err!("{}: failed to parse title", resp.url()));
+        }
+    }
+
     // vector to hold page content, which is progressively built from chunks of
     // downloaded data until a title is found (up to CHUNKS_MAX chunks)
     let mut body = Vec::new();