TL;DR Although in a story we use the past tense to denote things that are still true today, in this case that reading would involve an absurdity. The choice is thus simplified to whether or not the waters are still in this arrangement.
As Roger Vadim noted, Hebrew can omit the copula, i.e. the form of to be that serves to link a subject and its complement. Thus, the Hebrew reads:
וַיַּ֣עַשׂ אֱלֹהִים֮ אֶת־הָרָקִיעַ֒ וַיַּבְדֵּ֗ל בֵּ֤ין הַמַּ֙יִם֙ אֲשֶׁר֙ מִתַּ֣חַת לָרָקִ֔יעַ וּבֵ֣ין הַמַּ֔יִם אֲשֶׁ֖ר מֵעַ֣ל לָרָקִ֑יעַ וַֽיְהִי־כֵֽן
Which I would gloss thus:
God made the expanse1 [and] he [or it]2 divided the water3 that [is/was/will be] under the expanse from the water that [is/was/will be] above the expanse [and]4 it was thus
Those two [is/was/will be] blocks are where we must supply a copula between "water" and "below" or "above". The problem is that if we want to supply a copula, we have to infer the tense that it would be in if it were there. Also as Roger noted, Biblical Hebrew actually uses two aspects instead of tenses: perfect and imperfect. We usually translate the perfect as past tense and the imperfect as present or future, in both English and French. But these translations are only rough fits, and suggest things to us that may be inconsistent with what the respective aspect would have suggested to the original reader.
If we choose the past ("Il sépara les eaux qui étaient au-dessous de l'étendue d'avec les eaux qui étaient au-dessus de l'étendue"), we might do so because we reason that we use the past even for things that are still true today ("perennial facts"). But consider that this use of the past tense also implies that the thing was already true then. That is, that the water was already below and above the expanse before its creation. Since this seems unlikely (why create the expanse if they're already separated?), using the past tense from this motivation is unwise.
Hence, we're inclined to select the other interpretation of the past tense: that the waters were but are no longer in this arrangement, i.e. "qui étaient (autrefois) ...", the waters that used to be ... Some translators might favour this reading because they link the expanse with Gen. 7's "floodgates of heaven" that were opened for the great flood, releasing the waters that were "above the expanse" but are no longer.5
This leaves us with a more direct reason for why we might choose between the present and past: if you believe that the arrangement so described is intended to be the one still in effect, you select the present (because the past would suggest the wrong perennial fact), but if you believe that the arrangement came to an end, you select the past.
One remark is that though the copula is often omitted, it's actually supplied fairly commonly in the perfect. We might thus expect it to be supplied here if the perfect were intended.6
1 Etymologically, "something spread out".
2 No explicit subject is provided for this verb, but it's 3rd masc. sg., which could agree with God or the expanse.
3 The Hebrew word for "water" looks like a morphological plural, but there is no point in translating this as English's plural "waters" because it never occurs without this ending; it doesn't actually mark the plural. We do the same with "heavens", but we ignore this for "God" which also looks like the plural and should be "Gods" if we were consistent. This is a holdover from old translations and it still persists in "Bible-ese" in English and other European languages despite being the wrong translation.
4 The tense for these pasts is the wayyiqtol, which looks as though it starts with "and", but is used so often in non-conjunctive contexts that we must conclude it does not inherently bear that sense. Without "and", supply punctuation as needed.
5 An unnecessary harmonization, to my mind, and also one that misses the fact that though that was supposed to be the first time it rained, it has continued raining since then, suggesting that there is still water "above the expanse" in the Biblical cosmology.
6 For my money, this makes the present tense the better translation. But I do have to concede that I'm an amateur. :) And we must avoid the fallacy of denying the antecedent and saying that since when the copula is explicit it's in the perfect (P -> Q), therefore if it's not explicit it's not in the perfect (~P -> ~Q). This is not a valid deduction, even if there is probabalistic correlation.