fix: address unresolved Copilot review feedback from PRs #3, #5, #6

[KP2048]

Summary

You flagged that Copilot left review comments on recent PRs that never got addressed before merge. Checked #3, #4, #5, #6 — all four had Copilot reviews; #4's finding was already fixed before merge, the rest weren't. This PR fixes all of them.

PR #3 (fluid/energy storage)

• ArchieEnergyStorage.readSnapshot and its Serializer.deserialize now clamp capacity to non-negative and amount to 0..capacity. A malformed/stale snapshot could otherwise leave the storage over-capacity or negative, wedging insert()/extract() permanently.
• gui.md's ProgressBar example: value ?: 0 / smeltTicks.toFloat() — Elvis's low precedence means this divides 0 by smeltTicks, not the observed value, so a real value passes through undivided instead of becoming a 0..1 fraction. Parenthesized.
• Added BlockEntityNBTHolderTests.testItemFluidAndEnergyFieldsPersistMutations covering itemField/fluidField/energyField save/load round-tripping — the one field-kind trio the existing suite didn't touch.

PR #5 (docs-sync / GUI+networking KDoc)

• GuiGraphics.pose/scissor claimed to always restore pose/scissor state even if the block throws, but used no try/finally — an exception left the pose stack or scissor rect unbalanced, breaking subsequent rendering. Wrapped both (the IntRect scissor overload now just delegates to the primary one instead of duplicating the logic).
• Text's KDoc says "ARGB text color" but the renderer passed color.rgb, silently dropping any alpha the caller set (vanilla's drawString forces alpha=0 to opaque, so this always rendered fully opaque regardless of intent). Switched to color.argb so alpha is actually honored.
• NetworkChannel.configServerbound's KDoc said persistence happens in decodeDispatchData — it doesn't; that function only does the permission check, in-memory decode, and broadcast/reject. The handler registered by configServerbound is the only place that calls ConfigSpec.save(). Corrected.

PR #6 (changelog automation)

• release-notes.yaml's base-branch detection used git branch -r --contains, which matches any branch the tag is merely reachable from, not just one it's the tip of — non-deterministic if reachable from several. Switched to --points-at.
• pr-title-lint.yml used pull_request_target unnecessarily (no checkout/exec of fork code happens, so the base-repo context it grants isn't needed and only widens blast radius for future changes) — switched to pull_request, and secrets.GITHUB_TOKEN to the intended github.token context.

Test plan

• :common:compileKotlin clean (covers the gametest source set too — new test compiles)
• Reviewed each fix against the actual current code/behavior it touches, not just the review comment text

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.