#13856: update gelu_bw op documentation on bfloat8_b limitations

[KalaivaniMCW]

Ticket

Link to Github Issue #13856

Problem description

Since gelu_bw ops uses ttnn::exp in its implementation, beyond a certain input range the output is very a large number and will cause flushing of nearby values. bfloat8_b is a block format datatype with shared exponent for every 16 consecutive elements.
https://docs.tenstorrent.com/tt-metal/latest/ttnn/ttnn/tensor.html#limitation-of-bfloat8-b
For more details: link

What's changed

Updated docs on the limitations of using bfloat8_b dtype for the gelu_bw and bias_gelu_bw ops

Checklist

• Post commit CI passes
• Blackhole Post commit (if applicable)
• Model regression CI testing passes (if applicable)
• Device performance regression CI testing passes (if applicable)
• (For models and ops writers) Full new models tests passes
• New/Existing tests provide coverage for changes