Move attr-dict to right after op name. (iree-org#8373)

This is a follow up from iree-org#8365 which
updates the asm form. We are in favor of putting extra attributes to the
beginning.

Use `emitc.ptr`in VMToEmitC conversions (iree-org#8364)

* Add support for `emitc.ptr` type

* Use `emitc.ptr` in ListSetRefOpConversion

* Use `emitc.ptr` in ListSetOpConversion

* Use `emitc.ptr` in FailOpConversion

* Use `emitc.ptr` in GlobalLoadOpConversion

* Use `emitc.ptr` in GlobalStoreOpConversion

* Use `emitc.ptr` for `void*` type

* Use `emitc.ptr` for `void**` type

* Use `emitc.ptr` for `uint8_t*` type

* Use `emitc.ptr` for `iree_vm_stack_t*` type

* Use `emitc.ptr` for vm module pointer types

* Use `emitc.ptr` for `iree_vm_function_call_t*` type

* Use `emitc.ptr` for `iree_vm_execution_result_t*` type

* Use `emitc.ptr` for `iree_vm_function_t*` type

* Use `emitc.ptr` for `iree_vm_buffer_t*` type

* Use `emitc.ptr` for `iree_vm_module_state_t` pointer types

* Use `emitc.ptr` for `iree_vm_list_t` pointer types

* Use `emitc.ptr` for remaining pointer types

Refactors the remaining non-`iree_vm_ref_t` pointer types.

[CUDA] Enable async copy along with aggressive pipelining (iree-org#8342

)

This start using async copy ops for tensorcore pipeline which is enabled
for sm80+ targets.
This also enabled more aggressive pipelining in order to take advantage
of the async ops.
This adds a pass for multibuffer transformation needed for pipelining. This
change is under review upstream and once it lands the pass will only call the
upstream transformation.

Fix typo in host build directory name

Implements iree_hal_cuda_*_command_buffer_update_buffer. (iree-org#8334)

[CUDA] Refactor distribution to thread/warp (iree-org#8306)

Remove dependency to workload_per_wg attribute and clean up the code to
distribute fro block to thread/warp.

Update to hasVerifier form (NFC) (iree-org#8297)

* Update to hasVerifier form (NFC)

verifier form is deprecated and will be removed.

Revert malformed change to emscripten demo script (iree-org#8305)

The regex I used in iree-org#8302 neglected
the case of line continuations within the invocation of CMake (which
tbf, were only present in this experimental script). Reverting rather
than moving the arg to the right place because we agreed it's not
really useful in this non-CI script.

Partial revert of iree-org@0f80995367c3

Fix wrong variable name in iree_wait_until_impl for Apple platform (i…

…ree-org#8296)

Merge pull request iree-org#8284 from google/benvanik-timed-notification

Adding support for timeouts on iree_notification_t + fixes.