Fix eip1153_tstore from EF

[iinaki]

closes lambdaclass/ethereum_rust#621

Changes

• Moved the transient storage handling to the Journal.
• Fixed the EF EIP-1153 tests, which checked the how the transient storage was being handled. To do this I updated the codegen_tstore and the codegen_tload, and their corresponding syscalls.

During testing I noticed that the changes made to the handle of the transient storage were positive and that the reason of the failing of the tests relies on MLIR taking up too much stack space, which causes to violently overflow at a given point.
The test passes if run with RUST_MIN_STACK=83886080 make test-eth. This means that our implementation of tstore and tload is correct. I propose to keep skipping the test and leave the problem of optimizing memory usage with MLIR for further on, while merging this PR as the changes proposed are useful I believe.

[github-actions]

Benchmarking results

Benchmark for program factorial

Open benchmarks

Command	Mean [s]	Min [s]	Max [s]	Relative
evm_mlir_factorial	2.391 ± 0.275	2.068	2.608	1.00
revm_factorial	7.810 ± 1.288	6.951	10.264	3.27 ± 0.66

Benchmark for program fibonacci

Open benchmarks

Command	Mean [s]	Min [s]	Max [s]	Relative
evm_mlir_fibonacci	2.181 ± 0.279	1.914	2.448	1.00
revm_fibonacci	7.834 ± 0.789	7.430	9.401	3.59 ± 0.58

[belfortep]

Should TSTORE refund gas? Here in EVM Codes the only opcode that seems to trigger refunds is SSTORE

[iinaki]

Right, I believed it did refund gas but it doesn't. Here in the EIP says so too. Thanks for finding that out!