fix(process): validate usage snapshots and exit codes

[andrewtdiz]

Summary

Closes #3040
Closes #2927
Closes #3041
Closes #3049

• validate process.cpuUsage(previousValue) objects like Node, including array/non-object rejection, required numeric user/system, and finite non-negative counters
• add process.threadCpuUsage(previousValue) lowering/codegen/runtime support with the same previous-value validation and delta calculation
• align process.exit(code) validation for numeric strings, fractional values, NaN/Infinity, and non-number inputs
• validate process.setMaxListeners(n), preserve fractional/Infinity listener limits, and return the real process object for chaining

Why Batched

These issues all sit in node:process validation/readback behavior, share the same parity harness surface, and reuse the same runtime helpers for Node-style numeric/error formatting.

Tests Added

• test-parity/node-suite/process/cpu/cpu-usage-validation.ts
• test-parity/node-suite/process/cpu/thread-cpu-usage-previous.ts
• test-parity/node-suite/process/events/max-listeners-validation.ts
• test-parity/node-suite/process/methods/exit-validation.ts

Verification

Using Node v25.9.0 on PATH for parity expected output:

• ./run_parity_tests.sh --suite node-suite --module process --filter validation - 4 pass, 0 fail
• ./run_parity_tests.sh --suite node-suite --module process/cpu --filter thread-cpu-usage-previous - 1 pass, 0 fail
• ./run_parity_tests.sh --suite node-suite --module process/cpu - 5 pass, 0 fail
• ./run_parity_tests.sh --suite node-suite --module process/events - 8 pass, 0 fail
• ./run_parity_tests.sh --suite node-suite --module process/methods - 12 pass, 0 fail
• cargo check -p perry-runtime -p perry-hir -p perry-codegen -p perry
• cargo build --release --quiet -p perry -p perry-runtime -p perry-stdlib
• cargo fmt --all -- --check
• git diff --check
• ./scripts/check_file_size.sh
• jq empty test-parity/known_failures.json

Known Limitations

process.threadCpuUsage() keeps Perry's current per-thread CPU split behavior: available thread CPU time is reported as user with system: 0 on supported platforms. This PR adds previous-value support and validation around that existing measurement.

Non-Goals

• no process.hrtime() changes
• no UID/GID/umask validation changes
• no generic EventEmitter refactor
• no known_failures.json edits

[andrewtdiz]

Closing this as a duplicate of #3496: #3496. #3496 is the locally verified branch linked from the issue comments for this process validation/readback cut.

[andrewtdiz]

Closing this as a duplicate of #3496, which covers the same process validation/readback issue cluster and is the active PR for #3040, #2927, #3041, and #3049.