fix: PGAME replay buffer insertion

[manon-but-yes]

Re-write the insert() method of Replay Buffer to fix #71 and #70. This fix also avoids the error raised in #67.

I tested this new buffer on the HalfCheetah with an episode length of 1000 and batch-size of 100 and in my case, it seems to mainly improve the max-fitness reached by the algorithm. Here is the comparison between MAP-Elites and PGAME with the old and the new buffer insertion (the numbers at the end of the algorithms name are simply unique IDs for each algo, non-relevant to this comparison):

I left this PR as a draft because the Replay Buffer class is also used for the URL algorithms and it would require additional testing for these algorithms.