[FEATURE] Improve Error handling for semantic analysis

[TheGupta2012]

Feature Description

Error Emission

• Currently, the semantic analyser stops at the first error and emits out an error message to the user about the about location and the error.
• This might be good for a small number of semantic errors but for a lot of errors, the user experience would degrade. They would have to run the analyser as many times as the number of errors which is not desirable.
• Moreover, if the user has a very large number of errors, spitting out all errors at once could be confusing.
• A good feature would be to add an error threshold in the analyser. With this, we can keep a list of errors and as soon as the number of errors reach this predefined threshold, we stop and emit all of these errors.

Error Codes

• Adding unique identifiers to the messages would help locate the line of code quicker than parsing through the python traceback. Eg - OSA_ERR_123: <<error message>> ( OSA: openqasm semantic analyser )

Implementation (Optional)

OPENQASM 3.0;
array[int[32], 2, 2] my_arr = { {1,2}, {3,4}};
int[32] a = my_arr[0]; // this is semantically incorrect 

qubit q;
if( a == 4) { // what do we refer as "a" here? 
    for int[32] i in [:a] {
         rx(pi * i ) q;
    }
}

In the above example, do we treat a as undefined or invalid? What does "invalid" exactly mean here? How to continue semantic analysis in this erroneous state?

• What is a good threshold level for errors?