[WIP] Bump jax 0.7.2

[JerryChen97]

Context:
It seems that jax 0.7.2 has some "good" feature that we don't need const var anymore.

Description of the Change:

Benefits:

Possible Drawbacks:

Related GitHub Issues:

[JerryChen97]

temp skip until 0.7.2 is fully compatible with catalyst

[JerryChen97]

This part is also needed due to the new Literal in jax 0.7.2

[codecov]

Codecov Report

❌ Patch coverage is 92.04893% with 26 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.40%. Comparing base (4c31904) to head (7a578e6).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
...nnylane/transforms/optimization/cancel_inverses.py	66.66%	9 Missing ⚠️
pennylane/capture/jax_patches.py	92.92%	8 Missing ⚠️
pennylane/ops/functions/equal.py	81.81%	4 Missing ⚠️
pennylane/ops/qubit/observables.py	88.88%	2 Missing ⚠️
...templates/subroutines/qchem/all_singles_doubles.py	60.00%	2 Missing ⚠️
pennylane/capture/custom_primitives.py	97.36%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #8604      +/-   ##
==========================================
- Coverage   99.43%   96.40%   -3.03%     
==========================================
  Files         587      588       +1     
  Lines       61879    62162     +283     
==========================================
- Hits        61529    59927    -1602     
- Misses        350     2235    +1885     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[rniczh]

Somethings should also be updated:

1. capture/expand_transforms.py should convert the tuple back to their original type

def _tuple_to_slice(t):
    """Convert a tuple representation of a slice back to a slice object.

    JAX converts slice objects to tuples for hashability in jaxpr parameters.
    This function converts them back to slice objects for use with indexing.

    Args:
        t: Either a slice object (returned as-is) or a tuple (start, stop, step)

    Returns:
        slice: A slice object
    """
    if isinstance(t, tuple) and len(t) == 3:
        return slice(*t)
    return t

def _is_dict_like_tuple(t):
    """Checks if a tuple t is structured like a list of (key, value) pairs."""
    return isinstance(t, tuple) and all(isinstance(item, tuple) and len(item) == 2 for item in t)

def _tuple_to_dict(t):
    """
    Recursively converts JAX-hashable tuple representations back to dicts,
    and list-like tuples back to lists.

    Args:
        t: The item to convert. Can be a dict, a tuple, or a scalar.

    Returns:
        The converted dict, list, or the original scalar value.
    """

    if not isinstance(t, (dict, tuple, list)):
        return t

    if isinstance(t, dict):
        return {k: _tuple_to_dict(v) for k, v in t.items()}

    if isinstance(t, list):
        return [_tuple_to_dict(item) for item in t]

    if isinstance(t, tuple):

        # A. Dict-like tuple: Convert to dict, then recurse on values
        if _is_dict_like_tuple(t):
            # This handles the main (key, value) pair structure
            return {key: _tuple_to_dict(value) for key, value in t}

        # B. List-like tuple: Convert to list, then recurse on elements
        else:
            return [_tuple_to_dict(item) for item in t]

    return t

@ExpandTransformsInterpreter.register_primitive(_create_transform_primitive())
def _(
    self, *invals, inner_jaxpr, args_slice, consts_slice, targs_slice, tkwargs, transform
):  # pylint: disable=too-many-arguments
    args = invals[_tuple_to_slice(args_slice)]
    consts = invals[_tuple_to_slice(consts_slice)]
    targs = invals[_tuple_to_slice(targs_slice)]
    tkwargs = _tuple_to_dict(tkwargs)

    def wrapper(*inner_args):
        return copy(self).eval(inner_jaxpr, consts, *inner_args)

    jaxpr = jax.make_jaxpr(wrapper)(*args)
    jaxpr = transform.plxpr_transform(jaxpr.jaxpr, jaxpr.consts, targs, tkwargs, *args)
    return copy(self).eval(jaxpr.jaxpr, jaxpr.consts, *args)

2. templates/state_preparations/mottonen.py It's not numpy array compatible anymore. Probably just use qml.math.size to get the length of state_vetctor. But however, the Literal thing also affect the decision of the decomposition result in test_transform_merge_amplitude_embedding_workflow. Should also need to aware of this.

length = qml.math.size(state_vector)
global_phase = qml.math.sum(-1 * qml.math.angle(state_vector) / length)

3. op_math/sum.py should not use copy

self.queue[op_hash] = [coeff, summand]

And cannot compare tracer with bool:

if coeff == 1: