Windows: stop full process-table CPU sampling on every session (rebase of #13405)

[oz-by-warp]

Description

This is @npapano42's #13405 rebased onto the latest master (no content changes), so it can be reviewed against current master and the latest version of the review skills.

Why a new branch instead of force-pushing the original PR branch: #13405 is from a fork (npapano42/warp), and pushing the rebased commit there was rejected — a GitHub App cannot create or update .github/workflows/* without the workflows permission, and rebasing onto current master necessarily carries updated workflow files (e.g. .github/workflows/changelog_draft.yml). So the rebased commit is pushed to a branch in this repo instead.

Original description, from #13405:

On Windows, is_kaspersky_running() enumerated the entire system process table with per-process CPU cycle-time sampling on every terminal session bootstrap (every tab/pane/subshell/SSH session), with no cache — solely to check whether a process named avp (Kaspersky) exists.

On Windows, per-process CPU sampling makes the kernel issue NtQueryInformationProcess(ProcessCycleTime) for every process, and the full enumeration issues NtQuerySystemInformation(SystemProcessInformation). Both force KeFlushProcessWriteBuffers, which broadcasts a synchronous inter-processor interrupt to every logical core. Across the whole process table, in bursts of session inits, on high-core-count machines (worse with HVCI, where each IPI is a hypercall) this keeps all cores spinning at DISPATCH_LEVEL long enough to trip the DPC watchdog and bugcheck the system (DPC_WATCHDOG_VIOLATION 0x133).

Fix

1. app/src/system/info.rs — refresh_all_processes() now uses a dedicated name-only ProcessRefreshKind (all_processes_refresh_kind() → ProcessRefreshKind::nothing()) instead of the CPU+memory refresh_kind() used by the single-PID self-poll. Process names are still populated, so processes_by_name("avp") is unchanged.
2. app/src/util/windows.rs — is_kaspersky_running() is cached in a OnceLock<bool>, so the full sweep runs at most once per launch instead of on every session bootstrap.

Linked Issue

Fixes #13404. Supersedes #13405 (same single commit, rebased).

• The linked issue is labeled ready-to-spec or ready-to-implement.
• Where appropriate, screenshots or a short video of the implementation are included below. (N/A — no UI change.)

Testing

Original author's testing (Windows 11, Ryzen 9 5900X, HVCI on) is described in #13405, including the regression test app/src/system/info_tests.rs::all_processes_refresh_kind_does_not_sample_cpu_or_memory and a standalone sysinfo probe confirming names are still populated under the name-only refresh.

Verification of the rebase itself, on this branch at current master:

• Rebase applied cleanly with no conflicts; the diff versus master is byte-identical to the original commit's diff.

• ./script/format --check → clean.

• cargo clippy -p warp --all-targets --tests -- -D warnings → clean.

• cargo check -p warp --all-targets → clean.

• Running the unit test itself in this sandbox OOM'd while linking the warp lib test binary, so it was not re-executed here; it is a pure assertion over ProcessRefreshKind flags and type-checks under clippy --all-targets.

• I have manually tested my changes locally with ./script/run

Agent Mode

• Warp Agent Mode - This PR was created via Warp's AI Agent Mode

CHANGELOG-BUG-FIX: Windows — stopped enumerating the full process table with per-process CPU sampling on every terminal session, which could cause DPC_WATCHDOG_VIOLATION / high CPU on machines with many cores.

Conversation: https://app.warp.dev/conversation/3f7e68c2-5574-4bd2-8eb0-42c13931321f
Run: https://oz.warp.dev/runs/019ff261-4ae9-79b5-968a-2e3e41581388

This PR was generated with Oz.