PR #3200: Add support for float8_e4m3fnuz and float8_e5m2fnuz.

Imported from GitHub PR openxla/xla#3200

This adds support for the two FP8 types `float8_e4m3fnuz` and `float8_e5m2fnuz` to XLA similar to `float8_e4m3fn`, `float8_e4m3b11`, and `float8_e5m2`.
Copybara import of the project:

--
3b96f8fe219c1ac1bec5c4b99ff9c51148706981 by Jake Hall <[email protected]>:

Add support for float8_e4m3fnuz and float8_e5m2fnuz.

Merging this change closes #3200

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#3200 from jakeh-gc:fp8_fnuz 3b96f8fe219c1ac1bec5c4b99ff9c51148706981
PiperOrigin-RevId: 543802274

Author: jakeh-gc

tsl/platform/float8.h

@@ -25,7 +25,7 @@ using float8_e4m3b11fnuz = ml_dtypes::float8_e4m3b11fnuz;
 // Deprecated: old name for backward-compatibility only.
 using float8_e4m3b11 = float8_e4m3b11fnuz;
 using float8_e5m2 = ml_dtypes::float8_e5m2;
-using float8_e5m2funz = ml_dtypes::float8_e5m2fnuz;
+using float8_e5m2fnuz = ml_dtypes::float8_e5m2fnuz;
 }  // namespace tsl
 
 #endif  // TENSORFLOW_TSL_PLATFORM_FLOAT8_H_

tsl/protobuf/dnn.proto

@@ -19,6 +19,8 @@ enum DataType {
   kBF16 = 7;
   kF8E5M2 = 8;
   kF8E4M3FN = 9;
+  kF8E5M2FNUZ = 10;
+  kF8E4M3FNUZ = 11;
 }
 
 // Describes how a convolution input or output layer's data is formatted.