LibIPC: Encode spans of trivial arithmetic types more efficiently

For stuff like Span<u8>, we should obviously grow the message buffer
once, and then copy all the bytes in one go.

Author: awesomekling

AK/Span.h

@@ -125,6 +125,8 @@ class Span : public Detail::Span<T> {
 public:
     using Detail::Span<T>::Span;
 
+    using ElementType = T;
+
     constexpr Span() = default;
 
     [[nodiscard]] ALWAYS_INLINE constexpr T const* data() const { return this->m_values; }

Libraries/LibIPC/Encoder.h

@@ -139,6 +139,7 @@ template<>
 ErrorOr<void> encode(Encoder&, URL::BlobURLEntry::MediaSource const&);
 
 template<Concepts::Span T>
+requires(!IsArithmetic<typename T::ElementType>)
 ErrorOr<void> encode(Encoder& encoder, T const& span)
 {
     TRY(encoder.encode_size(span.size()));
@@ -149,6 +150,18 @@ ErrorOr<void> encode(Encoder& encoder, T const& span)
     return {};
 }
 
+template<Concepts::Span T>
+requires(IsArithmetic<typename T::ElementType>)
+ErrorOr<void> encode(Encoder& encoder, T const& span)
+{
+    TRY(encoder.encode_size(span.size()));
+
+    VERIFY(!Checked<size_t>::multiplication_would_overflow(span.size(), sizeof(typename T::ElementType)));
+    TRY(encoder.append(reinterpret_cast<u8 const*>(span.data()), span.size() * sizeof(typename T::ElementType)));
+
+    return {};
+}
+
 template<typename T, size_t N>
 ErrorOr<void> encode(Encoder& encoder, Array<T, N> const& array)
 {