[FR] Convert indices to more efficient versions

There's a few optimizations possible around indexing with an index_min_max we can utilize more. Because an index_min_max() has no guaranteed order (it can be a index_min_max(1, 5) or index_min_max(5, 1). we have to make a hard copy when taking a subset using them. But with index_min, index_max, and index_omni we have specializations that don't have to copy when taking subsets.

If a user writes

x[1:5] = y;

This get's generated as something like

assign(x, index_list(index_min_max(1, 5)), y)

but we know that's actually an index_max since the minimum index range in stan is 1. So if we see a 1:N we can transform the generated code to

assign(x, index_list(index_max(5)), y)

converting to index_min is a little more complicated. If a user writes

vector[N] x;
x[8:N] = y

This get's generated as something like

assign(x, index_list(index_min_max(8, N)), y)

but we know that N is the size of x so we could write

assign(x, index_list(index_min(8)), y)

Then finally it's possible to make 1:N min_max into an index_omni, which is effectivly a no-op

vector[N] x;
x[1:N] = y

This get's generated as something like

assign(x, index_list(index_min_max(1, N)), y)

but we know that N is the size of X and 1 is the smallest range value possible in stan so we could write

assign(x, index_list(index_omni()), y)

Start with the indexing specializations in src/stan/model/indexing/rvalue.hpp and trace how stanc3 generates index_min_max for the 1:N and 8:N examples. Verify generated C++ uses index_max, index_min, or index_omni where the bounds prove those forms equivalent, and confirm the resulting assignments preserve behavior.