Fix conv1d eager/graph mode inconsistency for zero-sized output

[ayushozha]

Summary

• Fixes Inconsistent behavior between tf.nn.conv1d eager execution and tf.function when output spatial dimension is zero #112371: tf.nn.conv1d with VALID padding silently produces a zero-sized tensor in eager mode when the kernel is larger than the input, while graph mode (tf.function) raises a ValueError
• Adds early validation in conv1d that checks if the output spatial dimension would be negative with VALID padding, raising a consistent ValueError in both execution modes
• The check only activates when static shape information is available, so it does not affect dynamic-shape scenarios

Test plan

• Verify that tf.nn.conv1d(tf.zeros([1, 2, 1]), tf.zeros([5, 1, 1]), stride=1, padding='VALID') now raises ValueError in eager mode
• Verify the same call under @tf.function still raises ValueError with a similar message
• Verify that valid conv1d calls (kernel <= input) are unaffected
• Verify that SAME padding is unaffected
• Verify dilated convolutions that produce negative output also raise the error

🤖 Generated with Claude Code

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.