202E; RIGHT-TO-LEFT OVERRIDE
   206A; INHIBIT SYMMETRIC SWAPPING
   206B; ACTIVATE SYMMETRIC SWAPPING
   206C; INHIBIT ARABIC FORM SHAPING
   206D; ACTIVATE ARABIC FORM SHAPING
   206E; NATIONAL DIGIT SHAPES

        *   `tf.math.argmin`/`tf.math.argmax` for input data type `tf.bool` on
            CPU.
        *   `tf.nn.gelu` op for output data type `tf.float32` and quantization
            on CPU.
    *   Add nominal support for unsigned 16-bit integer tensor types. Note that
        very few TFLite kernels support this type natively, so its use in mobile
        ML authoring is generally discouraged.

- Update kube-apiserver's priority & fairness work estimator such that 'max seats' is MIN(0.15 x nominalCL, nominalCL / handSize)
  
  This fixes a bug where clients with requests using hand size x max seats greater than the nominal concurrency limit can starve other requests in the same priority level. ([#118601](https://github.com/kubernetes/kubernetes/pull/118601), [@andrewsykim](https://github.com/andrewsykim)) [SIG API Machinery]

  // Limited priority levels in proportion to their NCS values:
  //
  // NominalCL(i)  = ceil( ServerCL * NCS(i) / sum_ncs )
  // sum_ncs = sum[limited priority level k] NCS(k)
  //
  // Bigger numbers mean a larger nominal concurrency limit,
  // at the expense of every other Limited priority level.
  // This field has a default value of 30.
  // +optional
  optional int32 nominalConcurrencyShares = 1;

    assertThat(stringPrep("\u200e")).isNull() // LEFT-TO-RIGHT MARK
    assertThat(stringPrep("\u206e")).isNull() // NATIONAL DIGIT SHAPES
    assertThat(stringPrep("\u206f")).isNull() // NOMINAL DIGIT SHAPES
  }

  @Test fun prohibitionTaggingCharacters() {
    assertThat(stringPrep("\udb40\udc01")).isNull() // Note that this is one code point.

  // (BorrowingCL) and is a limit on the total number of seats that this
  // level may borrow at any one time.
  // This field holds the ratio of that limit to the level's nominal
  // concurrency limit. When this field is non-nil, it must hold a
  // non-negative integer and the limit is calculated as follows.
  //
  // BorrowingCL(i) = round( NominalCL(i) * borrowingLimitPercent(i)/100.0 )
  //

  // (BorrowingCL) and is a limit on the total number of seats that this
  // level may borrow at any one time.
  // This field holds the ratio of that limit to the level's nominal
  // concurrency limit. When this field is non-nil, it must hold a
  // non-negative integer and the limit is calculated as follows.
  //
  // BorrowingCL(i) = round( NominalCL(i) * borrowingLimitPercent(i)/100.0 )
  //

  // (BorrowingCL) and is a limit on the total number of seats that this
  // level may borrow at any one time.
  // This field holds the ratio of that limit to the level's nominal
  // concurrency limit. When this field is non-nil, it must hold a
  // non-negative integer and the limit is calculated as follows.
  //
  // BorrowingCL(i) = round( NominalCL(i) * borrowingLimitPercent(i)/100.0 )
  //

- In the API Priority and Fairness feature, priority levels that are exempt from limitation can now be given a nominal and a lendable concurrency and their dispatching borrows from the concurrency limits of the other priority levels.  For details see https://github.com/kubernetes/enhancements/tree/master/keps/sig-api-machinery/1040-priority-and-fairness#dispatching...

- Update kube-apiserver's priority & fairness work estimator such that 'max seats' is MIN(0.15 x nominalCL, nominalCL / handSize)
  
  This fixes a bug where clients with requests using hand size x max seats greater than the nominal concurrency limit can starve other requests in the same priority level. ([#118601](https://github.com/kubernetes/kubernetes/pull/118601), [@andrewsykim](https://github.com/andrewsykim)) [SIG API Machinery]

## Dependencies