HRPOpt fails with current SciPy because _LINKAGE_METHODS was removed

[Samarthuday]

## Description

`HRPOpt.optimize()` fails with current SciPy versions because it accesses the private attribute `scipy.cluster.hierarchy._LINKAGE_METHODS`.

## Environment

- PyPortfolioOpt: 1.6.0
- Python: 3.14.6
- SciPy: 1.18.0
- macOS arm64

## Reproduction

```python
import pandas as pd
from pypfopt import HRPOpt

prices = pd.DataFrame(
    {
        "A": [100, 101, 99, 102, 103],
        "B": [100, 98, 101, 100, 104],
        "C": [100, 102, 103, 101, 105],
        "D": [100, 99, 100, 103, 102],
    }
)

returns = prices.pct_change().dropna()

hrp = HRPOpt(returns)
hrp.optimize()

Expected behavior

HRPOpt.optimize() should return portfolio weights for valid linkage methods.

Actual behavior

The call raises:

AttributeError:
module 'scipy.cluster.hierarchy' has no attribute '_LINKAGE_METHODS'

The error occurs during linkage-method validation before the portfolio is optimized.

Impact

HRP optimization fails with current SciPy versions, causing HRP-related tests to fail.

Pull request intention

I would like to submit a pull request that replaces this private SciPy API usage with a compatible validation approach and adds regression tests for the supported linkage methods.