Introduce a "transferable" algorithm for BufferSource

This is particularly helpful if several buffers are processed in a batch, as the validity of the operation can be determined up front, rather than during processing which could leave the buffers in a mix of states.

See discussion in webmachinelearning/webnn#351.

index.bs

@@ -9054,6 +9054,21 @@ a reference to the same object that the IDL value represents.
     1.  Return [$IsDetachedBuffer$](|jsArrayBuffer|).
 </div>
 
+<div algorithm>
+    A [=buffer source type=] instance |bufferSource| is
+    <dfn for="BufferSource" export>transferable</dfn> if the following steps return true:
+
+    1.  Let |jsArrayBuffer| be the result of [=converted to a JavaScript value|converting=]
+        |bufferSource| to a JavaScript value.
+    1.  If |jsArrayBuffer| has a \[[ViewedArrayBuffer]] internal slot, then set |jsArrayBuffer| to
+        |jsArrayBuffer|.\[[ViewedArrayBuffer]].
+    1.  If [$IsSharedArrayBuffer$](|jsArrayBuffer|) is true, then return false.
+    1.  If [$IsDetachedBuffer$](|jsArrayBuffer|) is true, then return false.
+    1.  If |jsArrayBuffer|.\[[ArrayBufferDetachKey]] is not <emu-val>undefined</emu-val>, then
+        return false.
+    1.  Return true.
+</div>
+
 <div algorithm>
     To <dfn for="ArrayBuffer" export>transfer</dfn> an {{ArrayBuffer}} |arrayBuffer|, optionally
     given a [=realm=] |targetRealm|: