Use .sort_unstable for speed

A stable sort is one where values that compare equal are in the same
order relative to each other in the output as they were in the input.
Thus, if input has values a0, b, and a1, with a0==a1, a0<b, a1<b, then
the output of a stable sort is guaranteed to be be [a0, a1, b]. For an
unstable sort, the output may also be [a1, a0, b]. This can matter if,
for example, the values are structs and only one field is used for
comparison.

On data where an unstable sort (.sort_unstable) works, it's bad to use
a stable sort (.sort), because it uses more CPU without affecting the
end result.

If the types of values in a vector are primitive types, as they are in
the cases changed by this commit, an unstable sort will do just fine.

Found by clippy lint stable_sort_primitive:
https://rust-lang.github.io/rust-clippy/master/index.html#stable_sort_primitive

1 files changed, 2 insertions, 2 deletions

diff --git a/openpgp/src/armor.rs b/openpgp/src/armor.rs
index cdb6b361..d009dacc 100644
--- a/openpgp/src/armor.rs
+++ b/openpgp/src/armor.rs
@@ -840,7 +840,7 @@ impl<'a> Reader<'a> {
                 // If there are no dashes at all, match on the BEGIN.
                 valid_start.push(b'B');
 
-                valid_start.sort();
+                valid_start.sort_unstable();
                 valid_start.dedup();
                 valid_start
             };
@@ -858,7 +858,7 @@ impl<'a> Reader<'a> {
                 // If there are no dashes at all, match on the BEGIN.
                 valid_start.push(b'B');
 
-                valid_start.sort();
+                valid_start.sort_unstable();
                 valid_start.dedup();
                 valid_start
             };