[keccak] Reduce Advice column usage in permutation circuit

[CPerezz]

Advice columns collapsing

StateBaseConversion - BaseConversion

We were previosly using 25 BaseConversionConfigs inside of a single
StateBaseConversion config.
Indeed, we can reach the same result and have 25 different regions for
each BaseConversionConfig without the need to have 25 different
instanciations.

This is important since each one of them adds 6 Advice columns and 2
Selector columns.

The after and before sum of columns for the permutation circuits shows a
decent reduction of almost a half of the columns in usage.

Before results for test_state_base_conversion:

Num advice columns: 151
Num fixed columns: 53
Num instance columns: 0
Num lookups: 25
Num gates: 50
num polys: 50
max_degree_poly: 3 - 
max_degree_lookup_input: 3 - lookup input 'Lookup i/o_coeff at Base conversion table'
max_degree_lookup: 1 - lookup table 'Lookup i/o_coeff at Base conversion table'

After results for test_state_base_conversion:

Num advice columns: 32
Num fixed columns: 5
Num instance columns: 0
Num lookups: 1
Num gates: 2
num polys: 2
max_degree_poly: 3 - 
max_degree_lookup_input: 3 - lookup input 'Lookup i/o_coeff at Base conversion table'
max_degree_lookup: 1 - lookup table 'Lookup i/o_coeff at Base conversion table'

Rho - LaneRotateConversionConfig

As a consequence of f507d16 and continuing the work on #489 this reduces
the RhoConfig to use a simple LaneRotateConversionConfig which gets
assigned 25 times witnesses generating 25 different Regions instead of
duplicating columns over and over.

This is important since each one of them adds 5 Advice columns, 2
Fixed columns and 2 Selector columns.

Before results for test_rho_gate:

Num advice columns: 156
Num fixed columns: 135
Num instance columns: 0
Num lookups: 52
Num gates: 52
num polys: 77
max_degree_poly: 3 - delta_acc === - coef * power_of_base
max_degree_lookup_input: 2 - lookup input 'b13 -> b9 table'
max_degree_lookup: 1 - lookup table 'b13 -> b9 table'

After results for test_rho_gate:

Num advice columns: 36
Num fixed columns: 15
Num instance columns: 0
Num lookups: 4
Num gates: 4
num polys: 5
max_degree_poly: 3 - delta_acc === - coef * power_of_base
max_degree_lookup_input: 2 - lookup input 'b13 -> b9 table'
max_degree_lookup: 1 - lookup table 'b13 -> b9 table'

[ChihChengLiang]

LGTM, Great improvement! Now we have much fewer columns overheads.
Added some nitpicks.

[ChihChengLiang]

We can still keep this lane and enable the copy for it.

[CPerezz]

I'm not sure I do follow. They'll be unused inside. So I thought the best was to enable it in the top level Config that instanciates LaneRotateConversionConfig.

[ChihChengLiang]

Enabling it in the top-level Config sounds good.

[CPerezz]

Final benchmarks before/after for keccak permutation after the column reduction/squash.
Before

test keccak_permutation::tests::bench_keccak_round has been running for over 60 seconds
Start:   EVM Proof generation with 19 rows
End: Setup generation with degree = 19 .........................................220.844s

After:

test keccak_permutation::tests::bench_keccak_round has been running for over 60 seconds
Start:   EVM Proof generation with 19 rows
End:     EVM Proof generation with 19 rows .........................................171.443s
test keccak_permutation::tests::bench_keccak_round ... ok