[WasmFS] Support file embedding

[kripken]

This practically worked since #16050, but the JS side needs to support a
missing parent, which is how the file packager emits it.

Fixes #16014

[kripken]

Minor issue, we had code paths that support both passing name,null and null,name to give just a name and not a pair of parent,name. Rather than support two ways of doing it, I standardized on one of them. This should fix the broken wasmfs test in the browser suite here.

[tlively]

It seems stranger to provide parent but not name than it would be to provide name but not parent. Would it be feasible to standardize in the other direction?

[kripken]

I think of it as name, null being "one param is passed, not two, so interpret the only one as the full path". We have similar things elsewhere in JS, where the dynamic nature makes it possible to consider arguments as "optional" that way. But I don't feel strongly, if you still disagree?

[tlively]

If it's consistent and idiomatic, then lgtm, but it definitely feels weird! Sometimes the parent parameter is actually a parent, but other times it's actually the name while the name parameter goes unused.

[kripken]

Hmm, I looked around a little for an example in our codebase, and can't find one atm... maybe it's not as common as I thought. I'll flip it around as you suggest.

[kripken]

Oh, I missed the example, which was just a few lines above this, in createPreloadedFile (as opposed to createDataFile which is modified here). That has the logic the first way I did it, and in fact tests break if it does not match between the two places. So I'll go back to the previous way...

[sbc100]

This seems odd to me too. The way I handled this in #16050 I guess made it work both wasy:

-      var path = name ? PATH.join2(typeof parent === 'string' ? parent : FS.getPath(parent), name) : parent;
+      var path = name;
+      if (parent) {
+        parent = typeof parent === 'string' ? parent : FS.getPath(parent);
+        path = name ? PATH.join2(parent, name) : parent;
+      }