Add benchmark work estimate for simple Cg solve

[upsj]

As a starting point and example for adding work estimates to kernels, this adds the necessary operations to all non-trivial kernels in a simple unpreconditioned Cg solve.

Example output for the simple-solver example in a debug build

Runtime summary

name	total	total (self)	count	avg	avg (self)	performance
total	2.5 ms	832.3 us	1	2.5 ms	832.3 us
apply(gko::solver::Cg)	1.6 ms	17.9 us	1	1.6 ms	17.9 us
iteration	1.5 ms	508.1 us	19	81.2 us	26.7 us
check(gko::stop::Combined)	305.9 us	91.0 us	20	15.3 us	4.6 us
apply(gko::matrix::Identity)	268.0 us	101.2 us	20	13.4 us	5.1 us
apply(gko::matrix::Csr<double, int>)	235.1 us	122.9 us	19	12.4 us	6.5 us
check(gko::stop::ResidualNorm)	193.1 us	148.3 us	20	9.7 us	7.4 us
copy(gko::matrix::Dense,gko::matrix::Dense)	166.8 us	142.7 us	20	8.3 us	7.1 us
csr::spmv	112.2 us	112.2 us	19	5.9 us	5.9 us	363.2 MB/s
advanced_apply(gko::matrix::Csr<double, int>)	67.1 us	38.8 us	2	33.5 us	19.4 us
dense::compute_conj_dot_dispatch	58.1 us	58.1 us	39	1.5 us	1.5 us	204.1 MB/s
generate(gko::solver::Cg::Factory)	50.3 us	50.3 us	1	50.3 us	50.3 us
cg::step_2	46.8 us	46.8 us	19	2.5 us	2.5 us	370.7 MB/s
cg::step_1	43.6 us	43.6 us	19	2.3 us	2.3 us	198.6 MB/s
dense::compute_norm2_dispatch	36.7 us	36.7 us	22	1.7 us	1.7 us	91.1 MB/s
csr::advanced_spmv	28.3 us	28.3 us	2	14.1 us	14.1 us	162.3 MB/s
dense::copy	24.1 us	24.1 us	20	1.2 us	1.2 us	252.0 MB/s
check(gko::stop::Iteration)	21.7 us	21.7 us	20	1.1 us	1.1 us
allocate	18.5 us	18.5 us	31	598.0 ns	598.0 ns
residual_norm::residual_norm	17.2 us	17.2 us	20	858.0 ns	858.0 ns
components::aos_to_soa	16.9 us	16.9 us	3	5.6 us	5.6 us
cg::initialize	15.0 us	15.0 us	1	15.0 us	15.0 us	70.7 MB/s
components::convert_idxs_to_ptrs	13.5 us	13.5 us	1	13.5 us	13.5 us
free	13.3 us	13.3 us	31	430.0 ns	430.0 ns
dense::fill	13.2 us	13.2 us	4	3.3 us	3.3 us	24.3 MB/s
components::fill_array	6.3 us	6.3 us	1	6.3 us	6.3 us
dense::fill_in_matrix_data	3.9 us	3.9 us	2	1.9 us	1.9 us

Overhead estimate 482.5 us

Work estimates available for 14.9 % of runtime