Fix false "Corrupted" images in the first-boot test

[Marketen]

Problem

On fresh v0.3.26 installs, the first-boot test reports every DAppNode image as corrupted, then says "Test completed successfully":

doing docker image integrity test...
Error response from daemon: invalid reference format: repository name (library/171MB) must be lowercase
\e[31mImage 171MB Corrupted!\n\e[0m
...
Test completed successfully. Press enter to continue

Seen on a real install (Debian, v0.3.26). The images are fine. dappnode_test_install.sh took image IDs from column 3 of docker images, but newer Docker versions changed the default table to IMAGE ID DISK USAGE CONTENT SIZE EXTRA. Column 3 is now the size, so the test ran docker save 171MB. The raw \e[31m is a separate bug: that echo lacked -e.

Fix

• Script: list and check images with explicit --format strings ({{.Repository}}:{{.Tag}}, {{.ID}}), so the output is the same on every Docker version, and fix the echo.
• Honest result: say "Test completed successfully" only when no image failed; otherwise list the corrupted images. The flow is unchanged: it still clears .firstboot and waits for Enter.
• Log: write the test output to /usr/src/dappnode/logs/dappnode_test_install.log. The test runs on its own console and used to leave no record, which also helps support.
• E2E: fail the ISO E2E when that log reports corrupted images or lacks the success line. Also fix an existing ! grep check in the harness, which can never fail under set -e.

Verification

• Ran the new integrity block against a stub docker with the new table: healthy images pass, a failing docker save flags exactly that image, and the old awk '{print $3}' reproduces 171MB / 2.08GB.
• The ISO E2E on this PR now checks the first-boot test's result on real Debian and Ubuntu installs.

🤖 Generated with Claude Code