Fix the logic for QuantizeConvert MatcherPass (#33271)
Quantized result nodes connected to an output port which is also
consumed by other node(s), gets unexpectedly converted into f32 type.
The TFLQuantizeConvert MatcherPass was matching for all patterns with a
TFLQuantize followed by a Convert op, and removing the Convert op by
setting the destination type directly to TFLQuantize. But, when another
node(eg. Result) is connected to such TFLQuantize node, the type change
causes mismatch for such nodes.
Changes done:
If the TFLQuantize node's output is connected to other nodes(apart from
the Convert), replace QuantizeConvert only for matching patterns.
### Details:
- *Fix the logic for QuantizeConvert MatcherPass*
### Tickets:
- *CVS-178652*
---------
Signed-off-by: Anoob Anto Kodankandath <anoob.anto.kodankandath@intel.com>
Co-authored-by: Maxim Vafin <maxim.vafin@intel.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
