cam_hal: PSRAM: forward-scan last dma node to find JPEG EOI reliably

[RubenKelevra]

PR requires the following PRs to be merged first, and thus contains their commits:

• fix jpeg soi check for psram_mode (DMA) #776
• fix jpeg eoi check for psram_mode (DMA) #777
• cam_hal: drop caches for psram_mode frames before hand-off to application #778

Issue

I've noticed that my previous implementation of the EOI search isn't
optimal, as it will not find the EOI, depending on where the image ends
in the last buffer due to padding.

Description

When psram_mode is active, cam_take() used to copy only 32 bytes
from PSRAM and scan backward for the JPEG EOI (FF D9). Because DMA
rounds the captured length up to a half-buffer, the true EOI often
falls earlier and was missed, leading to repeated "NO-EOI" resets.

This change:

• Drops the 32-byte SRAM copy; scan directly in PSRAM after cache
  invalidation.
• Searches the last DMA half-buffer plus 1 byte (marker_len-1),
  starting 1 byte before the final block to catch split EOIs.
• Scans forward so we pick the earliest EOI in the current tail and
  avoid stale markers from a larger previous frame.
• Uses a fast byte-hunt for 0xFF with a 2-byte verification for "FF D9".

SRAM mode keeps the simple backward check (we only accept EOI at offset
0 there), so no extra complexity or perf work is added.

Net effect: robust EOI detection without garbage tails and fewer
spurious "NO-EOI" resets in PSRAM captures.

Testing

Does fix some crashes, but not everything in psram_mode with JPEGs.

It will now reliably fetch one frame and then the frame capture stalls somewhere until the watchdog kills it (after 4 seconds) and the device reboots. I'm investigating.

---

Checklist

Before submitting a Pull Request, please ensure the following:

• 🚨 This PR does not introduce breaking changes.
• All CI checks (GH Actions) pass.
• Documentation is updated as needed.
• Tests are updated or added as necessary.
• Code is well-commented, especially in complex areas.
• Git history is clean — commits are squashed to the minimum necessary.

[me-no-dev]

@RubenKelevra please fix the conflicts and errors

[RubenKelevra]

Minor fix added, that dma_node_buffer_size is used instead of dma_half_buffer_size as search window, as the DMA operations are done with the former. No functional change in most situations, as both are usually the same for JPEG operations.

[RubenKelevra]

Thanks @me-no-dev !