TensorFlow frontend Softplus converter still emits native softplus op (fp16 overflow risk on ANE)

## Problem

The TensorFlow/Keras frontend's Softplus converter at `coremltools/converters/mil/frontend/tensorflow/ops.py:2149` still emits the native `mb.softplus` MIL op:

python
def Softplus(context, node):
...
x = mb.softplus(x=x, name=node.name) # native op → fp16/ANE overflow

This is the same native op that overflows in fp16 on Apple Neural Engine at `x ≈ 10.4`, causing output collapse to 0.

## Context

PR #2725 fixes this for the **PyTorch frontend** by replacing `mb.softplus()` with the numerically stable decomposition `max(x,0) + log(1+exp(-|x|))`. The TF frontend was intentionally left out of scope since both original issues (#2687, #2359) originated from PyTorch models.

## Proposed Fix

Apply the same `_stable_softplus_mil(x)` helper to the TF Softplus converter. Alternatively, implement this at the MIL graph-rewrite pass level so all frontends are covered (the DRY fix, as suggested by @ChinChangYang in the PR #2725 review).

## Related
- PR #2725 (PyTorch frontend fix)
- Issue #2687 (original overflow report)
- Issue #2359 (Mish fp16 errors)

Start at coremltools/converters/mil/frontend/tensorflow/ops.py:2149 and compare the TensorFlow Softplus converter with the stable PyTorch change in PR #2725. Apply the stable Softplus handling to the TensorFlow path, or confirm a graph-rewrite approach covers it, and verify that the converter no longer emits the native softplus operation for this case.

Written by the indexing model from the issue text.