feat: enable auto-batching passes

[avik-pal]

For the reduction case. With these new passes:

julia> Reactant.XLA.cost_analysis(fn1)
HloCostAnalysisProperties
-------------------------
FLOPS: 49920.0
Transcendentals: 0.0
Bytes Accessed: 150.938 KiB
Optimal Seconds: 0.0
Utilization: 0.0
Operand 0 Utilization: 2.0
Operand 1 Utilization: 1.0
Operand 0 Bytes Accessed: 120.234 KiB
Operand 1 Bytes Accessed: 240 bytes
Output Root Bytes Accessed: 30.469 KiB
Reserved 0: 0.0

Without

julia> Reactant.XLA.cost_analysis(fn2)
HloCostAnalysisProperties
-------------------------
FLOPS: 199.0
Transcendentals: 0.0
Bytes Accessed: 210.563 KiB
Optimal Seconds: 0.0
Utilization: 0.0
Operand 0 Utilization: 12.0
Operand 1 Utilization: 5.0
Operand 0 Bytes Accessed: 105.023 KiB
Operand 1 Bytes Accessed: 436 bytes
Output Root Bytes Accessed: 105.122 KiB
Reserved 0: 0.0