feat!: non-exhaustive public error enums, build-time E2E caps (0.11.0)

[JustinKovacich]

Closes #144. Closes #147.

The first of the batch raised against the 0.10.0 surface. These two travel together because both land as 0.11.0 and both touch the "what can a consumer configure or match on" surface; #145, #146 and #148 follow separately.

#[non_exhaustive] on the public error enums (#144) — breaking

Applied to client::Error, server::Error, e2e::Error, protocol::Error, and protocol::sd::Error. transport::TransportError already had it — so this makes the treatment uniform rather than introducing it.

Six enums, not the five the issue named: protocol::sd::Error turned up in the sweep.

client::Error's doc carried a stability note asserting the opposite, and honestly spelling out the cost it was accepting — "any new variant added here is a breaking change." That note is rewritten rather than deleted, because the reasoning is still worth stating, just with the conclusion inverted.

Build-time E2E caps (#147) — additive

E2E_REGISTRY_CAP and E2E_RX_STATE_CAP now go through from_env_or, matching SERVICE_REGISTRY_CAP, the server caps, and the bare-metal runtime. Defaults unchanged at 32 and 64.

Verified all three paths, not just the happy one:

Build	Result
SIMPLE_SOMEIP_E2E_RX_STATE_CAP=128	builds
SIMPLE_SOMEIP_E2E_RX_STATE_CAP=100	fails — E2E_RX_STATE_CAP must be a power of two for heapless::FnvIndexMap
SIMPLE_SOMEIP_E2E_REGISTRY_CAP=abc	fails — non-digit guard

The power-of-two const _ assertions were previously defensive against a future edit; they are now load-bearing against user input, which is worth being deliberate about — a non-power-of-two fails the build rather than silently mis-sizing the map.

Two things reviewers should look at

from_env_or loses its #[cfg(any(feature = "server", feature = "client"))] gate. That gate existed because every caller lived in one of those modules and it would otherwise be dead code. e2e is compiled unconditionally, so once the caps use it the gate breaks --no-default-features outright — and there is no longer any configuration in which the function is dead. Caught by building that config; it is in the verification list below for that reason.

The 0.11.0 bump is two-part. simple-someip-embassy-net carries version = "0.11" next to its path = ".." dependency, so bumping only the root Cargo.toml fails resolution for the whole workspace. Same shape as the version bump in 0.10.0.

Also corrected in passing: from_env_or's non-digit assert said SIMPLE_SOMEIP_MAX_*, which only covers one of the two var families (MAX_OFFERS/MAX_SUBS vs the *_CAP set). It now says SIMPLE_SOMEIP_* and the doc names both.

Verification

• cargo clippy -- -D warnings -D clippy::pedantic on all five lanes CI runs: the alloc/host workspace lane, --no-default-features, and the three bare_metal combinations.
• cargo test --no-default-features — CI's test lane — green.
• cargo doc --no-deps with RUSTDOCFLAGS=-D warnings for client, server,bare_metal, and the alloc set. This is what caught an intra-doc link I had first written into the new e2e docs pointing at client::SERVICE_REGISTRY_CAP, which is behind a feature the e2e module does not have.
• cargo fmt --all -- --check clean.
• --no-default-features builds — the configuration the cfg-gate change is about.

Unrelated, but noting it since it is visible if you run the full suite: cargo test --workspace --features $ALLOC_FEATURES fails 12 subscriber tests on unmodified main as well. That combination is not one CI's test lane runs. Not touched by this branch, and not investigated here.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

@@            Coverage Diff             @@
##             main     #150      +/-   ##
==========================================
- Coverage   81.26%   81.25%   -0.01%     
==========================================
  Files          46       46              
  Lines       15470    15470              
==========================================
- Hits        12571    12570       -1     
- Misses       2899     2900       +1     

Files with missing lines	Coverage Δ
src/client/error.rs	100.00% <ø> (ø)
src/e2e/registry.rs	96.58% <ø> (ø)
src/lib.rs	0.00% <ø> (ø)
src/server/error.rs	100.00% <ø> (ø)

... and 1 file with indirect coverage changes