Duplicate reaction from different libraries make it into the model when only one of the reactions is ThirdBody

Bug Description

We ran RMG with primaryH2O2 and JetSurF2.0, and got the following reactions in the core:


! Reaction index: Chemkin #267; RMG #4198
! Library reaction: JetSurF2.0
! Flux pairs: H2O(7), H2O(7); H(3), H2(4); H(3), H2(4); 
H(3)+H(3)+H2O(7)=H2(4)+H2O(7)                       5.624000e+19 -1.250    0.000    

! Reaction index: Chemkin #251; RMG #9
! Library reaction: primaryH2O2
! Flux pairs: H(3), H2(4); H(3), H2(4); 
H(3)+H(3)+M=H2(4)+M                                 7.000e+17 -1.000    0.000    
Ar/0.00/ H2(4)/0.00/ H(3)/0.00/ H2O(7)/14.30/

Cantera complained that these are duplicate reactions:

CanteraError: 
*******************************************************************************
InputFileError thrown by Kinetics::checkDuplicates:
Undeclared duplicate reactions detected:
Reaction 264: 2 H(3) + M <=> H2(4) + M
Reaction 248: 2 H(3) + H2O(7) <=> H2(4) + H2O(7)
*******************************************************************************

Cantera is "correct" since the general reaction H(3)+H(3)+M=H2(4)+M also considers H2O as the collider (it's efficiency is not set to 0.

How To Reproduce

Run a model with both libraries that involves oxidation.

Expected Behavior

We expect that if a general reaction was added from one library (the +M reaction), then a specific reaction with a specific collider from another library will not be added. If the reverse occurs it is more problematic, yet we should (?) assume that each library is self-consistent and does not include only a single specific collider reaction without the "mother reaction".

ReactionMechanismGenerator / RMG-Py