Add support for operations needed for well-known transformers e.g. Se…

…gment Anything, Stable Diffusion, etc. (#478)

* almost done.

* Done.

* tri -> triangular

* Fix Bikeshed link errors.

* Minor formatting change.

* Minor code fix. No unsqueeze needed here since the affine params must already be in shape,

* Addressing much of the PR feedback. Still more to go.

* Addressing more PR feedback.

* More fixes.

* Fix dropping the mixin.

* Separate out argMin/Max. More 'type' -> 'dataType`.

* add 'be'

* Add dfn for MLOperand.shape() and .dataType(). Fix argMin/Max return type.

* Fix `where` emulation code.

* Fix some wording to clarify the return shape of `gather`.

* Fix layernorm input shape typo.

* Acknowledge Dwayne's contributions.

* Wording adjustment.

* }