tests: refresh duplicate-basename comments for NEXTEST_BINARY_ID shim

[cargo-affected-bot]

The duplicate_target_names.rs and lib_bin_collision.rs test files document the binary-resolution mechanism in their docstrings and inline comments. Those mechanisms changed in PR #14 (shim: source binary_id from NEXTEST_BINARY_ID, drop the binary map), but the comments still describe the prior approaches:

• The (package, target, kind) triple-based lookup that PR Disambiguate lib+bin same-basename collisions in shim marker probe #14 replaced.
• The pre-run listing that built a binary_map — no longer exists.
• The "byte-marker" / source-path fingerprinting fallback that depended on debug info — also no longer exists; the env variable carries the unambiguous <package>::<target> id directly.

The test scenarios themselves are still valid regression guards (duplicate basenames + stripped debuginfo + partial rebuild are real-world shapes), but the explanatory comments mislead future readers about how the current shim resolves binaries.

This PR refreshes only the comments — no code or assertion changes, no behavioural shift.

[cargo-affected-bot]

In-diff comments are accurate against src/shim.rs. Two adjacent lines in the same files were left referring to the old shim and fit the same refresh:

• tests/functional/duplicate_target_names.rs#L60-L61 — the cargo_affected_stripped helper docstring still attributes the RUSTFLAGS='-C debuginfo=0' to defeating "the old marker fallback". The sibling helper in lib_bin_collision.rs:88-90 already carries the up-to-date rationale (mirror the stripped-binary CI environment that originally tripped the bug).
• tests/functional/lib_bin_collision.rs#L120-L121 — the assertion message reads "marker probe must disambiguate lib+bin", but the marker probe is gone; the !stderr.contains("basename fallback ambiguous") check itself is still a valid regression guard.

Both are out-of-diff so I haven't posted inline suggestions; happy to push a follow-up commit if you'd like them folded in.

[cargo-affected-bot]

Added one follow-up commit: the assert panic message at line 121 still said "marker probe must disambiguate lib+bin", referencing the same defunct mechanism this PR removes elsewhere. Updated it to "NEXTEST_BINARY_ID must disambiguate lib+bin" to match shim.rs ("no marker probe needed"). The nearby basename fallback ambiguous string check is left as-is — it's a regression guard against reintroducing the old probe path.

[cargo-affected-bot]

Pushed one more follow-up: the cargo_affected_stripped helper docstring (line 60-61) was the last spot still attributing RUSTFLAGS='-C debuginfo=0' to defeating the "old marker fallback". That mechanism is gone — refreshed it to say debuginfo=0 mirrors the stripped-binary CI environment that originally tripped the duplicate-basename bug, matching the sibling helper in lib_bin_collision.rs and this file's own header. No code or assertion changes. With this, both files are free of stale references to the pre-NEXTEST_BINARY_ID mechanisms.