Enable HPU Fused SDPA for Qwen3-VL vision attention using attention masks

[slokesha]

Qwen3-VL vision attention is updated to use FusedSDPA.apply directly when the query sequence length is within the supported fused range (q_len ≤ 65536).
This removes the per-block Q/K/V attention loop and enables the optimized HPU fused SDPA kernel for vision attention.

The change aligns Qwen3-VL with the optimized path already used by Qwen2.5-VL on Gaudi, improving efficiency while preserving identical model outputs.

[github-actions]

🚧 CI Blocked

The main CI workflow was not started for the following reason:

This is a Draft PR. Please mark it as 'Ready for Review' to trigger the CI.

[github-actions]

🚧 CI Blocked

The main CI workflow was not started for the following reason:

This is a Draft PR. Please mark it as 'Ready for Review' to trigger the CI.

[slokesha]

Performance:
FusedSDPA enabled -
Serving Benchmark Result
Successful requests: 100
Failed requests: 0
Maximum request concurrency: 8
Benchmark duration (s): 336.64
Total input tokens: 5280
Total generated tokens: 11827
Request throughput (req/s): 0.30
Output token throughput (tok/s): 35.13
Peak output token throughput (tok/s): 120.00
Peak concurrent requests: 15.00
Total token throughput (tok/s): 50.82
---------------Time to First Token----------------
Mean TTFT (ms): 10186.15
Median TTFT (ms): 10396.45
P99 TTFT (ms): 28860.68
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms): 138.33
Median TPOT (ms): 124.15
P99 TPOT (ms): 246.55
---------------Inter-token Latency----------------
Mean ITL (ms): 139.06
Median ITL (ms): 71.97
P99 ITL (ms): 2633.77

Main -
Serving Benchmark Result
Successful requests: 100
Failed requests: 0
Maximum request concurrency: 8
Benchmark duration (s): 334.12
Total input tokens: 5280
Total generated tokens: 11855
Request throughput (req/s): 0.30
Output token throughput (tok/s): 35.48
Peak output token throughput (tok/s): 113.00
Peak concurrent requests: 15.00
Total token throughput (tok/s): 51.28
---------------Time to First Token----------------
Mean TTFT (ms): 10806.36
Median TTFT (ms): 11990.35
P99 TTFT (ms): 22304.62
-----Time per Output Token (excl. 1st token)------
Mean TPOT (ms): 131.59
Median TPOT (ms): 111.62
P99 TPOT (ms): 231.90
---------------Inter-token Latency----------------
Mean ITL (ms): 131.67
Median ITL (ms): 73.13
P99 ITL (ms): 2470.87

[github-actions]

🚧 CI Blocked

The main CI workflow was not started for the following reason:

Your branch is behind the base branch. Please merge or rebase to get the latest changes.

[github-actions]

🚧 CI Blocked

The main CI workflow was not started for the following reason:

This is a Draft PR. Please mark it as 'Ready for Review' to trigger the CI.