[Feature] 集成 TurboQuant KV Cache 压缩

[sunghajung6688]

将 TurboQuant V3（纯 MSE 模式）集成到 pypto-serving 的 KV Cache 管理中，支持在线压缩/解压缩历史 token，减少 KV Cache 内存占用。

[gemini-code-assist]

Code Review

This pull request introduces TurboQuant KV cache quantization to optimize memory usage during prefill and decode phases. The implementation includes a new turboquant module featuring MSE-optimal compression using random rotation and Lloyd-Max quantization. Feedback focuses on critical performance bottlenecks: the current compression logic re-processes the entire sequence prefix in each step, leading to $O(N^2)$ complexity; the manual Python loops used for writing to paged cache should be replaced with vectorized advanced indexing; and several constant tensor allocations and PDF calculations should be pre-computed or vectorized to reduce overhead.

[gemini-code-assist]

当前实现在每次调用时都会重新压缩整个序列前缀（从 token 0 到 tokens_used - residual_window）。在 decode 循环中调用时，这将导致相对于序列长度的 $O(N^2)$ 复杂度，随着上下文增长，性能将严重下降。建议实现增量压缩，仅处理刚刚超出 residual_window 的新 token。此外，调用的 read_context 方法内部也存在类似的循环读取瓶颈，建议一并优化。

[gemini-code-assist]

在 Python 循环中遍历 token 并写入分页缓存（paged cache）非常低效，特别是对于设备张量（NPU/GPU）。应使用 PyTorch 的高级索引（advanced indexing）进行向量化，以在单个操作中完成更新。

Suggested change

	for row in range(keys.shape[0]):
	token_index = start_token_index + row
	page_idx = token_index // pool.page_size
	offset = token_index % pool.page_size
	physical_page = alloc.page_ids[page_idx]
	pool.key_pages[layer_idx, physical_page, :, offset, :] = keys[row].to(cache_dtype)
	pool.value_pages[layer_idx, physical_page, :, offset, :] = values[row].to(cache_dtype)
	token_indices = torch.arange(start_token_index, start_token_index + keys.shape[0], device=keys.device)
	page_indices = token_indices // pool.page_size
	offsets = token_indices % pool.page_size
	physical_pages = torch.tensor(alloc.page_ids, device=keys.device)[page_indices]
	pool.key_pages[layer_idx, physical_pages, :, offsets, :] = keys.to(cache_dtype)
	pool.value_pages[layer_idx, physical_pages, :, offsets, :] = values.to(cache_dtype)

[gemini-code-assist]

idx_powers 是一个常量张量。在每次 compress 调用时创建它是低效的。建议在 __init__ 中根据 bits 预先计算并存储为 buffer，以减少冗余分配和 H2D 拷贝。

[gemini-code-assist]

高斯 PDF 计算目前在 Python 列表推导式中执行，对于 2048 个样本来说较慢。可以使用 PyTorch 操作轻松实现向量化，提高初始化速度。

            xs = torch.linspace(a, b, n_samples)
            pdf_vals = (1.0 / (math.sqrt(2 * math.pi) * sigma)) * torch.exp(-xs**2 / (2 * sigma**2))
            weighted = xs * pdf_vals