mtmd: add batching support for internvl

[ngxson]

Overview

This PR is mostly serves as a demo on how to adapt model's cgraph from non-batching to batching support.

InternVL cgraph seems to be very simple, a good starting point

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: no

[ngxson]

Tested with the new script from #24778 --> batching up to 3 images OK:

[vision] OK:   ggml-org/InternVL2_5-1B-GGUF:Q8_0
0.17.024.664 I encoding mtmd batch, n_chunks = 3 (done = 145, total = 161)
0.17.224.927 I mtmd batch encoding done in 200 ms
0.17.502.190 I encoding mtmd batch, n_chunks = 3 (done = 149, total = 161)
0.17.703.914 I mtmd batch encoding done in 201 ms
0.17.982.876 I encoding mtmd batch, n_chunks = 3 (done = 153, total = 161)
0.18.184.286 I mtmd batch encoding done in 202 ms
0.18.473.397 I encoding mtmd batch, n_chunks = 3 (done = 157, total = 161)
0.18.675.175 I mtmd batch encoding done in 202 ms

The publisher's name on the newspaper is 'KAPPERZASZAK'.
[vision] OK:   ggml-org/InternVL3-1B-Instruct-GGUF:Q8_0
0.17.143.834 I encoding mtmd batch, n_chunks = 3 (done = 145, total = 161)
0.17.346.677 I mtmd batch encoding done in 202 ms
0.17.617.924 I encoding mtmd batch, n_chunks = 3 (done = 149, total = 161)
0.17.823.393 I mtmd batch encoding done in 205 ms
0.18.101.736 I encoding mtmd batch, n_chunks = 3 (done = 153, total = 161)
0.18.307.555 I mtmd batch encoding done in 206 ms
0.18.597.972 I encoding mtmd batch, n_chunks = 3 (done = 157, total = 161)
0.18.803.980 I mtmd batch encoding done in 206 ms

The newspaper is called 'Kaperejszak'.