fix: align .text section start to 8 bytes

[kosmar]

Summary

• _stext = ADDR(.start_block) + SIZEOF(.start_block) places .text at 0x1000013c, which is not a multiple of the section’s current 8-byte alignment requirement. rust-lld has been warning about this on every build (“address of section .text is not a multiple of alignment”).
• Fix: ALIGN(..., 8) pads 4 bytes after .start_block so .text starts at 0x10000140 and satisfies the ELF output-section alignment invariant.
• Note: this corrects the aggregate section-header alignment; it does not by itself prove that the earlier flaky app-task spawn crashes were caused by that misalignment (shifting code addresses can also mask a separate layout-sensitive bug).

Found while debugging dense preset pushes with the preset editor (#602). Host-side pacing there reduces USB pressure; this change clears the linker alignment warning and keeps .text properly aligned.

Test plan

• Build firmware: rust-lld alignment warning is gone
• Flash to hardware and push a dense layout (10+ apps incl. multiple instances of the same app) via the configurator

[chmanie]

The linker-script change itself looks correct: ALIGN(..., 8) moves .text from 0x1000013c to 0x10000140 and satisfies its current 8-byte output-section alignment requirement. One requested wording change: please remove or soften the claim in memory.x and the PR description that this alignment issue caused the app-task spawn crashes. The original linker output still aligned individual input sections as needed; this patch fixes the aggregate ELF section-header invariant, but shifting code addresses can also mask a separate layout-sensitive bug. Suggested comment:

/* Move .text after the boot info and satisfy its current 8-byte
 * output-section alignment requirement. */
_stext = ALIGN(ADDR(.start_block) + SIZEOF(.start_block), 8);

With that wording adjustment, I have no issue with the alignment implementation.

[chmanie]

Thank you for the fix @kosmar! If you allow edits by maintainers for this PR, we can also take care of rebasing and testing this properly.

[kosmar]

@chmanie Thanks — updated the memory.x comment and softened the PR description accordingly. Maintainer edits are enabled, so feel free to rebase and test whenever you have a chance.

[ArthurGibert]

Reviewed and empirically verified. Built main pre-fix and reproduced the exact rust-lld alignment warning this PR fixes; built this branch and confirmed the warning is gone, .text lands 8-aligned (confirmed via readelf), and the resulting release ELF still converts to a valid UF2. No correctness issues.

One minor, non-blocking note: the ALIGN(..., 8) value is hardcoded with nothing asserting it matches .text's actual required alignment. If a future toolchain/dependency bump raises that requirement, the same warning would silently reappear — it's a linker warning, not a rustc one, so it isn't caught by -D warnings or any CI gate. Not worth blocking on.

[ArthurGibert]

Approved on my side. @chmanie let me know

[chmanie]

@ArthurGibert yeah fine by me