[FEATURE] Add noise options for tactile sensors: hysteresis, dead taxels, probe gain

[Milotrince]

Description

• Schmitt-trigger hysteresis on ContactProbe with separate contact_threshold / release_threshold
• Viscoelastic hysteresis on ContactDepthProbe, KinematicTaxel, ProximityTaxel, ElastomerTaxel
• Spatial crosstalk on KinematicTaxel, share FFT code with ElastomerTaxel
• Per-taxel probe_gain applied to measured contact depth, with optional resample-on-reset (probe_gain_resample_range).
• Dead taxels — dead_taxel_probability + dead_taxel_value_range, resampled per scene.reset()
• Allow 0 radius probes to be used like padding, so that we can use FFT acceleration for non perfect grid patterns

Related Issue

Motivation and Context

• More realistic noise parameters to match real tactile sensors

How Has This Been / Can This Be Tested?

Added tests to test_sensors.py

Screenshots (if appropriate):

noisy.mp4

Checklist:

• I read the CONTRIBUTING document.
• I followed the Submitting Code Changes section of CONTRIBUTING document.
• I tagged the title correctly (including BUG FIX/FEATURE/MISC/BREAKING)
• I updated the documentation accordingly or no change is needed.
• I tested my changes and added instructions on how to test it for reviewers.
• I have added tests to cover my changes.
• All new and existing tests passed.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f6ff7d64a3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Sample dead-taxel replacement per output channel

The dead-taxel replacement values are sampled with shape (n_envs, total_n_probes) and later broadcast to cache columns via cache_col_probe_idx, so multi-channel taxels (e.g., force/torque xyz) get the same scalar in every channel for a dead probe. This makes dead outputs artificially correlated and does not match the documented per-channel dead-value behavior, so dead KinematicTaxel/ProximityTaxel readings lose channel independence after each reset.

Useful? React with 👍 / 👎.