build: trim dev-profile debuginfo to shrink target/ bloat

[mukeshblackhat]

Summary

First step toward addressing #257 (target/ bloat). Trims dev-profile debuginfo workspace-wide:

• [profile.dev] debug = "line-tables-only" — keep stack traces + panic line numbers, drop DWARF type/variable tables
• [profile.dev.package."*"] debug = false — dependencies shed all debuginfo (we never debug into them)

Release profile is untouched on purpose — CI / release builds produce identical artifacts.

Measured impact

Reproducible on a fresh cargo clean, full workspace, on main vs this branch:

Stage	Command	BEFORE	AFTER	Saved	%
1	cargo build --workspace	3.7 G	2.4 G	1.3 G	−35%
2	+ cargo test --workspace --no-run	6.4 G	4.3 G	2.1 G	−33%
3	+ cargo build --workspace --release	8.0 G	6.0 G	2.0 G	−25%

Release portion is 1.6 G in both runs — the savings come entirely from target/debug/, exactly as intended.

Build times: roughly identical (within noise; debug-info emission is a small share of total compile time).

Why this doesn't break anything

debug = "line-tables-only" controls what debuginfo the compiler emits, not codegen. Same machine code, same symbols, same linking. Stack traces still pinpoint our source lines; panic messages still print the right file/line. The only thing lost is the ability to inspect local variables of a dependency in a debugger — which nobody does for tokio / axum / rust-embed etc.

Test plan

Anyone can reproduce the comparison locally:

# === BEFORE (on main) ===
git checkout main && git pull
cargo clean
cargo build --workspace                   ; du -sh target  # ~3.7 G
cargo test --workspace --no-run           ; du -sh target  # ~6.4 G
cargo build --workspace --release         ; du -sh target  # ~8.0 G

# === AFTER (this branch) ===
git checkout investigate/target-bloat-257
cargo clean
cargo build --workspace                   ; du -sh target  # ~2.4 G  ✅
cargo test --workspace --no-run           ; du -sh target  # ~4.3 G  ✅
cargo build --workspace --release         ; du -sh target  # ~6.0 G  ✅

# correctness
make lint   # ✅ zero warnings
make test   # ✅ all green

• make lint — zero warnings
• make test — all green
• cargo build --workspace --release — succeeds, output unchanged (release profile not touched)
• Disk usage measured BEFORE and AFTER on a clean target/ — numbers above

Out of scope (follow-up PRs, tracked in #257)

• B. Narrow make watch from cargo build --workspace to cargo build -p coast-daemon
• C. Extend make prune to also remove coast-service-dev-{cargo,target,docker} Docker volumes
• D. CONTRIBUTING.md note on disk hygiene
• E. Remove stale /target-cov entry in .gitignore (no code writes there)

Closes — none yet; refs #257, parent issue stays open until B–E land.

---

Reviewer note: per CONTRIBUTING.md, a Loom walkthrough of the change is required — I'll attach one once the PR is reviewed for correctness, since this is a config-only change with measured numbers above.