Optimization: Replace np.empty_like(a) with np.empty(a.shape, dtype=a.dtype) after pd_to_np conversion

[SaFE-APIOpt]

ta_cn/ta_cn/ewm_nb.py

Line 148 in a569a61

out = np.empty_like(a)

Hi, thank you for your great work!

In the following snippet:

a = pd_to_np(a)
out = np.empty_like(a)

after converting a pandas object to a NumPy array, the shape and dtype of a are already known and reliable. While np.empty_like(a) is valid and convenient, it incurs minor overhead due to internal type-checking and metadata copying.

A faster and equally correct alternative is:
out = np.empty(a.shape, dtype=a.dtype)
This avoids unnecessary introspection and speeds up array initialization, particularly in performance-critical loops or batch processing routines.

[wukan1986]

Recommend using expr_codegen+polars_ta for more convenient use