[BUG FIX] Fix taichi debug mode errors and warnings

[Libero0809]

Description

• Add ti.i32 and ti.u32 explicit typing to avoid overflow.
• Change expand_bits function to avoid taichi compiler warning in debug modes.
• Fix bugs in indexing.

Related Issue

Resolves Genesis-Embodied-AI/Genesis#

Motivation and Context

Now the code can be run in taichi debug mode without outputting tons of warning messages.

How Has This Been / Can This Be Tested?

Tested with code

import taichi as ti
import time

@ti.func
def expand1(x: ti.u32) -> ti.u32:
    x = (x | (x << 16)) & ti.u32(0xFF0000FF)
    x = (x | (x <<  8)) & ti.u32(0x0F00F00F)
    x = (x | (x <<  4)) & ti.u32(0xC30C30C3)
    x = (x | (x <<  2)) & ti.u32(0x49249249)
    return x

@ti.func
def expand2(x: ti.u32) -> ti.u32:
    x = (x * ti.u32(0x00010001)) & ti.u32(0xFF0000FF)
    x = (x * ti.u32(0x00000101)) & ti.u32(0x0F00F00F)
    x = (x * ti.u32(0x00000011)) & ti.u32(0xC30C30C3)
    x = (x * ti.u32(0x00000005)) & ti.u32(0x49249249)
    return x

@ti.func
def expand3(x: ti.u32) -> ti.u32:
    x = (x * ti.u32(0x00010001)) & ti.u32(0xFF0000FF)
    x = (x | ((x & 0x00FFFFFF) <<  8)) & 0x0F00F00F
    x = (x * ti.u32(0x00000011)) & ti.u32(0xC30C30C3)
    x = (x * ti.u32(0x00000005)) & ti.u32(0x49249249)
    return x

@ti.kernel
def test1(n_test: ti.i32):
    for i, j in ti.ndrange(n_test, 1024):
        data1[i, j] = expand1(j)

@ti.kernel
def test2(n_test: ti.i32):
    for i, j in ti.ndrange(n_test, 1024):
        data2[i, j] = expand2(j)

@ti.kernel
def test3(n_test: ti.i32):
    for i, j in ti.ndrange(n_test, 1024):
        data3[i, j] = expand3(j)

if __name__ == "__main__":
    ti.init(arch=ti.gpu, debug=False)
    n_test = 100000
    n_rounds = 10
    data1 = ti.field(ti.u32, shape=(n_test, 1024))
    data2 = ti.field(ti.u32, shape=(n_test, 1024))
    data3 = ti.field(ti.u32, shape=(n_test, 1024))
    for _ in range(n_rounds):
        data1.fill(0)
        start = time.time()
        test1(n_test)
        print("expand1 time:", time.time() - start)
        data2.fill(0)
        start = time.time()
        test2(n_test)
        print("expand2 time:", time.time() - start)
        data3.fill(0)
        start = time.time()
        test3(n_test)
        print("expand3 time:", time.time() - start)
        assert (data1.to_numpy() == data2.to_numpy()).all()
        assert (data1.to_numpy() == data3.to_numpy()).all()

Negligible difference:
expand1 time: 0.01114034652709961
expand2 time: 0.01214289665222168
expand3 time: 0.012060403823852539
expand1 time: 3.123283386230469e-05
expand2 time: 1.7642974853515625e-05
expand3 time: 1.6927719116210938e-05
expand1 time: 1.3589859008789062e-05
expand2 time: 1.2636184692382812e-05
expand3 time: 1.3589859008789062e-05
expand1 time: 1.239776611328125e-05
expand2 time: 1.3113021850585938e-05
expand3 time: 1.1920928955078125e-05
expand1 time: 1.2874603271484375e-05
expand2 time: 6.103515625e-05
expand3 time: 1.3589859008789062e-05
expand1 time: 1.239776611328125e-05
expand2 time: 1.2159347534179688e-05
expand3 time: 1.1920928955078125e-05
expand1 time: 1.2159347534179688e-05
expand2 time: 1.2159347534179688e-05
expand3 time: 1.1920928955078125e-05
expand1 time: 1.1920928955078125e-05
expand2 time: 1.1682510375976562e-05
expand3 time: 1.1920928955078125e-05
expand1 time: 1.6450881958007812e-05
expand2 time: 1.239776611328125e-05
expand3 time: 1.3589859008789062e-05
expand1 time: 1.33514404296875e-05
expand2 time: 1.430511474609375e-05
expand3 time: 1.2636184692382812e-05

Screenshots (if appropriate):

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.