image/probe plots misaligned with data for deep recordings

**Describe the bug**
For deep insertions in the DR dataset, I've found that sometimes the image/probe plots can be misaligned with the data (see below for example). It appears the upper and lower banks data has been swapped. For DR data, surface finding recordings and main recordings are separate assets, which is likely part of the bug.

**To Reproduce**
Steps to reproduce the behavior:
1. Examine the following experiment in the IBL GUI: 796848_2025-07-14_16-02-52

**Screenshots**
In this example, it's clear that the upper bank LFP correlation and RMS plots are being plotted on the bottom. This is particularly apparent for this insertion, as though it had a surface finding recording, the probe was only inserted 3 mm (i.e. the entire upper bank is out of the brain):

**Desktop (please complete the following information):**
- generated with latest released conversion/IBL GUI capsules (v2.0)

Reproduce by examining experiment 796848_2025-07-14_16-02-52 in the IBL GUI using the latest released conversion/IBL GUI capsules v2.0. Start by tracing how DR data with separate surface-finding and main recording assets maps upper and lower bank LFP correlation/RMS plots. Done means the image/probe plots align with the data for this deep insertion and the banks are no longer swapped.

由索引模型根據 Issue 內容生成。